ABSTRACT
Title of dissertation: COMPETING OR COLLABORATING
SIBLINGS? AN INVESTIGATION OF
THE RELATIONSHIP BETWEEN
INDUSTRIAL AND TRADE POLICIES
Gunjan Sharma, Doctor of Philosophy, 2006
Dissertation directed by: Professor Roger Betancourt
Department of Economics
This dissertation investigates the relationship between industrial and trade
policies and their impact on firm-level incentives to become more productive. In
Chapter two we use a two-sector growth model and show that the impact of a rise
in competition in the intermediate goods sector (that invests in quality-enhancing
technology) is sensitive to market structure in the final goods sector. We find that
more competition in the intermediate goods sector (due to industrial policy reform)
can lead to rising investment in technology and hence rising productivity. Further
we find that industries that face more competition domestically can perform better
in the face of foreign competition. That is, there may be strategic complementarities
between industrial deregulation and trade reform. We also find that a rise in compe-
tition in the final goods sector can affect investment incentives in the intermediate
goods sector and hence affect productivity. This study highlights the importance
of market structure assumptions in growth models. The third chapter tests predic-
tions from chapter two using two unique data sets. We use the industrial licensing
regime in India (operating from the 1950s onwards) and its gradual relaxation dur-
ing the 1980s and 1990s to test whether industrial de-regulation that leads to more
competition domestically, affects firm-level productivity. To our knowledge, ours is
the only detailed data set on Indian industrial policy. Our firm-level data for the
period 1980-94 is a census of firms in India and has been rarely used in literature.
We also use the interesting chronology of reforms in India (industrial de-regulation
in the 1980s and trade reforms in 1991) to test whether industries that faced more
competition domestically tend to perform better when facing foreign competition.
Our identification strategy uses an important institutional feature of Indian policy.
Firms with assets below a certain defined rupee threshold were exempt from licens-
ing requirements. This institutional feature provides us within-industry variation
that allows us to identify the interaction between de-licensing and exemption status.
We find that industrial de-regulation during the 1980s led to a significant rise in firm
productivity. Further preliminary results suggest that there exists a strategic com-
plementarity relationship between industrial and trade policies?industries and firms
that were de-licensed tend to perform better vis productivity after trade liberaliza-
tion. Our results are robust to the inclusion of a wide variety of firm and industry
fixed effects and controls for policies other than de-licensing that may affect produc-
tivity. This chapter contributes to the literature by being the only detailed empirical
analysis of the industrial licensing regime in India, especially the de-licensing that
took place during the 1980s and by providing evidence of the crucial link between
trade and industrial de-regulation.
COMPETING OR COLLABORATING SIBLINGS?
AN INVESTIGATION OF THE RELATIONSHIP BETWEEN
INDUSTRIAL AND TRADE POLICIES
by
Gunjan Sharma
Dissertation submitted to the Faculty of the Graduate School of the
University of Maryland, College Park in partial fulfillment
of the requirements for the degree of
Doctor of Philosophy
2006
Advisory Committee:
Professor Roger Betancourt, Chair/Advisor
Professor Nuno Limao
Professor Deborah Minehart
Professor Christopher Mckelvey
Professor Sonalde Desai
c? Copyright by
Gunjan Sharma
2006
DEDICATION
For Nani and Daddy, Mom and Papa, and Gaurang. Your faith, love and
support have made this possible.
ii
ACKNOWLEDGMENTS
I am grateful to a lot of people whose help and support made this thesis
possible and who made the past five years of my life an unbelievable experience.
First and foremost, I am very thankful to my advisor, Dr. Roger Betancourt,
for agreeing to work with me, for keeping patience and faith in me and in my
ideas, for providing invaluable guidance-both academic as well as practical, for never
refusing to meet with me and for providing practical, moral and personal support
through the various stages of this thesis. I have learnt an immense amount from
him and I know I will never forget those lessons.
I would also like to thank Dr. Nuno Limao for his theoretical ideas and em-
pirical expertise, his help and his attention to detail. This thesis has genuinely
benefitted from his guidance. Thanks are due to Dr. Deborah Minehart for her
belief in my vision, her extraordinary suggestions and her detailed, pain-staking
comments on my manuscripts. I would also like to thank Dr. Sonalde Desai and
Dr. Christopher Mckelvey for agreeing to serve on my thesis committee and for
sparing their invaluable time reviewing the manuscript.
The major part of this thesis would have been impossible without the requisite
data. In this regard I am highly grateful to Mr. B.S. Minhas and Dr. Adarsh Kishore
who helped me to locate the data and find ways to procure it; my father, Mr. M.L.
Sharma and my mother Dr. Vinita Sharma, who spent a lot of time following up
iii
on my case in New Delhi; and the library staff at the National Council for Applied
Economic Research, New Delhi, who generously made available their facilities and
their time to me. I would also like to thank the many officials in the Ministry of
Program Planning and Implementation, the Ministry of Industry and the Industry
of Commerce who spared me their valuable time and whose frank conversations
truly heightened my understanding of the issues.
This thesis has benefitted immensely from my conversations with Dr. Arvind
Panagariya and Dr. Aaditya Mattoo. Both of them, with their detailed, in-depth
knowledge of the Indian growth experience and policy experiments, provided in-
valuable ideas and suggestions. I also thank all the brown-bag participants at the
Department of Economics, University of Maryland for their frank comments and
suggestions.
I would like to give a special mention to everyone at the Maryland Population
Research Center for taking me in, allowing me to use their facilities and expertise
and drink their tea. In particular, Sarbartha Bandhopadhyay (Bandy) responded
immediately and practically to my cry for help and has been a constant source of
advice and help. Without his support and that of the Pop Center, there was almost
no chance of this thesis being completed on time.
I owe my deepest thanks to my family members. My grandparents-Prof. A. N.
Parashar and Mrs. Vidya Parashar- always encouraged me to follow my academic
dreams and their love, courage and guidance is with me even though they are not
anymore. My parents have always stood by me and have been beacons of hope and
encouragement during the best and the worst parts of the past five years. They
iv
have encouraged, cajoled, counseled and scolded me at all the right times. Thanks
to my brother, Gaurang, for providing comedic relief through-out and to my best
friends, Malini and Sudarshan, for helping me keep things in perspective and for
continuously reminding me of a wider world lurking outside my thesis.
And the one last thing that this thesis required was plenty of shoulders to
cry on. My band of girls-Isabel, Yyannu and Helena; my erst-while roommates-
Archana, Rachana and Marilyn; my friends-Sudarshan, Rahul, Anuj, Srikant, Nitish.
They have been my friends, confidantes and pressure valves. They have always
provided me with an avenue to discuss my woes, with companionship, with dinner
and with an enriching graduate student experience.
Lastly, I thank God for allowing me to finish my thesis and for helping me to
realize that I truly love what I do.
v
TABLE OF CONTENTS
List of Tables viii
List of Figures ix
1 Introduction 1
2 Strategic Complementarities between Trade and Industrial Policies: A The-
oretical Investigation 8
2.1 Literature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2 Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2.1 Finals Goods Sector . . . . . . . . . . . . . . . . . . . . . . . 13
2.2.2 Intermediate Goods Sector . . . . . . . . . . . . . . . . . . . . 14
2.2.3 Investment by Intermediate Goods Sector:
Production of Quality . . . . . . . . . . . . . . . . . . . . . . 15
2.2.4 Foreign Entry into the Intermediate Goods Sector . . . . . . . 17
2.3 Market Structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.3.1 The Benchmark Case: Competitive FG, Monopolist IG . . . . 20
2.3.2 Monopoly in the FG Sector . . . . . . . . . . . . . . . . . . . 26
2.3.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.4 Robustness Checks . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
2.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3 Competing or Collaborating Siblings? Industrial and Trade Policies in India 43
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.2 Literature Survey . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.3 Background on Indian Industrial Policy . . . . . . . . . . . . . . . . . 53
3.4 Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
3.5 Identification Strategy and Estimation Equation . . . . . . . . . . . . 60
3.5.1 Estimation Equation . . . . . . . . . . . . . . . . . . . . . . . 60
3.5.2 Identification Problem and Strategy . . . . . . . . . . . . . . . 61
3.5.3 Exogeneity of Exemption Status . . . . . . . . . . . . . . . . . 69
3.5.4 Comparability of non-equivalent groups . . . . . . . . . . . . . 75
3.5.5 Controls . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
3.6 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
3.6.1 Main Results . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
3.6.2 Controlling for firm-level un-observables . . . . . . . . . . . . 83
3.7 The Relationship between Trade and Industrial Reforms . . . . . . . 88
3.7.1 Does Deregulation Matter? . . . . . . . . . . . . . . . . . . . . 88
3.7.2 Testing Predictions from Theory . . . . . . . . . . . . . . . . . 90
3.8 Conclusions and Future Work . . . . . . . . . . . . . . . . . . . . . . 97
vi
A Appendices to Chapter 2 110
A.1 Proof of Equations 2.20 and 2.21 . . . . . . . . . . . . . . . . . . . . 110
A.2 Cost Minimization Exercise of FG Monopolist . . . . . . . . . . . . . 114
A.3 Proof of Propositions 1 and 2 . . . . . . . . . . . . . . . . . . . . . . 116
A.4 Proofs of Propositions 3 and 4 . . . . . . . . . . . . . . . . . . . . . . 120
A.5 Regulated Monopolist in the Final Goods Sector . . . . . . . . . . . . 121
B Appendices to Chapter 3 125
B.1 Background on Industrial Licensing in India . . . . . . . . . . . . . . 125
B.1.1 The Procedures . . . . . . . . . . . . . . . . . . . . . . . . . . 125
B.1.2 Scope of Licensing . . . . . . . . . . . . . . . . . . . . . . . . 128
B.1.3 Implementation of the License Raj . . . . . . . . . . . . . . . 130
B.2 Behavior of Non-licensed Firms . . . . . . . . . . . . . . . . . . . . . 137
B.2.1 Nature of Asset Limits . . . . . . . . . . . . . . . . . . . . . . 137
B.2.2 Distribution of Firms Around the Threshold . . . . . . . . . . 138
B.2.3 Capital-Labor Ratio . . . . . . . . . . . . . . . . . . . . . . . 140
B.3 Specification Test using Special Industries . . . . . . . . . . . . . . . 144
B.4 Specification Test using Various Thresholds for Exemption . . . . . . 147
B.5 Comparability of Exempt and Not Exempt Firms . . . . . . . . . . . 151
vii
LIST OF TABLES
2.1 Productivity of IG Firm . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.1 Percentage of Output and Value Added De-licensed in each year. . . . 101
3.2 Characteristics of the data . . . . . . . . . . . . . . . . . . . . . . . . 101
3.3 Summary Statistics for All Factories. . . . . . . . . . . . . . . . . . . 102
3.4 Average Annual Rates of Growth of Assets(%). . . . . . . . . . . . . 102
3.5 Baseline Results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
3.6 Robustness Checks. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
3.7 Results for Pseudo-Panel of firms. . . . . . . . . . . . . . . . . . . . . 104
3.8 Impact of Trade Reforms. . . . . . . . . . . . . . . . . . . . . . . . . 105
3.9 Year Effects in Trade Regression . . . . . . . . . . . . . . . . . . . . . 106
3.10 Impact of Trade Reforms on Advanced Industries. . . . . . . . . . . . 107
B.1 Utilization Rates for Exempt firms . . . . . . . . . . . . . . . . . . . 138
B.2 Distribution of Firms near the Threshold for Licensing . . . . . . . . 141
B.3 Distribution of Firms near the Threshold for Licensing . . . . . . . . 143
B.4 Falsification Test using Special Industries. . . . . . . . . . . . . . . . 146
B.5 Coefficient on Interaction for Various Thresholds . . . . . . . . . . . . 150
B.6 Separate Regressions for Sub-samples . . . . . . . . . . . . . . . . . . 152
B.7 Coefficient on Interaction for Various Percentiles of Firms . . . . . . . 159
B.8 Separate Regressions for the 40th Percentile . . . . . . . . . . . . . . 160
viii
LIST OF FIGURES
3.1 Distribution of Assets of Exempt firms: Assets?Rs.50 million. . . . . 107
3.2 Distribution of Assets of Exempt firms: Rs.50 million?Assets>Rs.4.5
million. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
3.3 Distribution of Assets of Exempt firms: Assets?Rs.4.5 million. . . . . 108
3.4 Distribution of Productivity in the Pseudo-panel . . . . . . . . . . . . 109
3.5 Distribution of Assets of Exempt Firms in the Pseudo-panel. . . . . . 109
A.1 Supply curve of Constrained Monopolist. . . . . . . . . . . . . . . . . 123
B.1 Distribution of Capital-Labor Ratio over Exempt Firms . . . . . . . . 142
B.2 Coefficient on Interaction and Confidence Intervals . . . . . . . . . . 149
B.3 Trends in Productivity-Exempt and Not Exempt Firms . . . . . . . . 152
B.4 Trends in Productivity over Industries . . . . . . . . . . . . . . . . . 154
B.5 Trends in Productivity over States . . . . . . . . . . . . . . . . . . . 155
B.6 Trends in Productivity over Ownership Structure . . . . . . . . . . . 155
B.7 Trends in Productivity over Organizational Structure . . . . . . . . . 156
ix
Chapter 1
Introduction
The interaction between trade liberalization and domestic industrial policy reform
is of great importance in both trade and development literature. Recent evidence
about the positive impact of trade liberalization (Tybout (2000), Epifani (2003))
on productivity, investment and welfare as well as the fact that several large and
developing economies have joined the mainstream of the world economy have demon-
strated that the economic and welfare implications of international trade are large
and can not be ignored. However, these very high stakes have led to a debate in
literature. On the one hand, some believe that competition from imports enhances
the incentives of the firm to upgrade their technology (see, for example, Dollar and
Kraay (2001), Frankel and Romer (1999), Sachs and Warner (1995)). On the other
hand, others believe that these incentives are affected by domestic institutions and
policies and trade reform will not be beneficial unless domestic polices are reformed
as well (Rodrik and Rodriguez (2001)). That is, domestic institutions and poli-
cies can be used to achieve a superior economic outcome after trade liberalization.
This issue is particularly important with regard to countries like India and China
that had, or still have, very controlled industrial and trade policies. These policies
created many distortions in these economies and it is of great interest to analyze?
both theoretically as well as empirically? the implications of these distortions for
1
the benefits from trade reforms.
The second chapter of this dissertation discusses market structure assumptions
in standard models of economic growth and demonstrates that these assumptions
are important in determining firm-level incentives to invest in quality-enhancing
technology. We are particularly interested in the cases like that of India with regard
to the kinds of industrial and trade policies that were in place. Since gaining inde-
pendence in the 1940s, India followed a policy of close control of private enterprize
in the manufacturing sector. The implementation of this control was via a licens-
ing regime, which meant that each and every firm in each manufacturing industry
needed to take permission from the government to enter or continue production.
That is, entry into manufacturing industry was not free (i.e. not determined by
market forces) and the threat of potential entry was low as well (since not every
application submitted for entry into an industry would be accepted by the author-
ities). This led to the creation of artificial monopolies and oligopolies in almost all
industries in the manufacturing sector.
This aspect of the licensing regime implies that the assumption of perfect com-
petition is not valid for any sector of the Indian economy. Thus theoretical models
that are used to generate predictions for India (for example, Aghion et al. (2003b),
Aghion et al. (2004)) must take into account the distortions created by industrial li-
censing. The second chapter of this dissertation tries to capture significant features
of industrial licensing in India in a theoretical model and in this way provides a
richer, more realistic analysis of the effects of reforms- both domestic and external-
on the incentive of firms to invest in quality enhancing technology.
2
The theory outlined in Chapter 2 demonstrates that a rise in competition in
the sector that invests in quality-enhancing technological progress will, under certain
conditions, lead to a rise in technological investment in that sector. This result is
different from the standard Schumpeterian hypothesis that a rise in competition
should, unambiguously, reduce firm-level incentives to invest in technology since
the ability of the firm to appropriate surplus from investment falls as competition
rises. Further, a rise in competition in downstream sectors (that do not invest
in technology) also leads to a rise in the investment incentives of the sectors that
invest. That is, a firm?s incentives to invest are affected not only by the degree of
competition in its own industry but also in other industries. Finally, we find that
under certain conditions, trade liberalization and domestic competition are strategic
complements. That is, a rise in domestic competition raises the marginal investment
response of the intermediate goods firm to a rise in the threat of foreign entry.
The third chapter of this dissertation empirically tests important implications
from Chapter 2. Our case study-India-is particularly interesting and relevant for
two reasons. Firstly, as mentioned earlier a very rigid and stern industrial licensing
regime was in operation for 40 years. Entry to into manufacturing industry was not
free (i.e. not based on market forces) and the threat of potential entry was low as
well. This in turn lead to monopolistic distortions in almost all sectors. The second
feature that makes India a good case study for the relationship between trade and
industrial reforms is the interesting chronology of reforms- industrial de-regulation
in the 1980s and trade reforms in the 1990s. This chronology of reforms allows us to
distinguish between the two types of reforms and to assess the relationship between
3
them. The results of our empirical exercise provide several important insights into
previously unexplored aspects of industrial licensing in India.
Licensing impacted a firm along two important dimensions. The workings of
the regime affected both the ability as well as the incentives of a firm to invest in
productivity. There were controls on the amount of output, the location of the plant,
the technology used etc that were conditions printed on the license document and
these directly controlled the ability of a firm to become productive. On the other
hand, the regime implicitly controlled entry into an industry and hence, affected
firm incentives to become productive. When an industry was de-licensed, both these
controls were removed simultaneously. In this context, our identification strategy
allows us to capture the impact of the relaxation of both direct and indirect controls
on firms. That is, we can measure the impact of de-licensing on a firm?s productivity
because of relaxation of microeconomic constraints on the firm as well as the impact
on productivity due to greater competition and a higher threat of potential entry
faced by a firm.
Our identification strategy uses the distinction between two types of firms
within an industry? small firms that were exempt from licensing provisions and large
firms that were not exempt from licensing provisions. The main difference between
these two types of firms was that Exempt firms did not need a license to operate
and hence were not obliged to fulfill microeconomic conditions on output, location,
technology etc. On the other hand, not exempt firms were required to obtain a
license to enter as well as fulfill onerous conditions. Technically this distinction
meant that there was free entry into the ranks of exempt firms in an industry and
4
that all entrants would like to enter as exempt firms. But this was unlikely to be
true. This is because the other accoutrements of the ?License-Quota-Permit Raj?
that operated in India meant that a manufacturing license was just a part of a whole
bundle of permissions and licenses that a firm needed to operate. Thus there were
implicit entry barriers into the ranks of exempt firms and explicit entry barriers into
the ranks of not exempt firms. This means that de-licensing of an industry reduced
entry barriers for both of exempt as well as not exempt firms and that our empirical
strategy catches these effects.
Our empirical study demonstrates that de-licensing of Indian industry (that
took place is a piece-meal manner during the 1980s followed by a big push in 1991)
lead to a rise in productivity of the firms that were affected by it. Thus the in-
tuition provided to us by the theory in Chapter 2 is borne out by data. In the
presence of market structure distortions in all sectors, firms that face a more com-
petitive environment domestically will respond by improving their productivity. It
is to be noted that de-licensing could impact the productivity of a firm by several
mechanisms other than that of a rise in competition. Our empirical application of
results from Chapter 2 does not address the issue of mechanism but provides highly
suggestive evidence.
Further, firms in industries that were deregulated prior to the trade reform
episode in 1991 performed much better after the trade reforms. Again, the strategic
complementarity result postulated in Chapter 2 is borne out by the Indian expe-
rience. This finding has an important policy implications for developed as well as
developing countries? it might be economically more efficient to target domestic dis-
5
tortions in order to make domestic firms more competitive in the world market. In
our case study, the domestic distortion was due to a policy of licensing. However, a
wide variety of other policies like production subsidies, entry barriers into industry,
financial aid to ailing domestic firms etc can be analyzed in the framework proposed
by this thesis.
This dissertation contributes to the theoretical and empirical literature in three
ways. Firstly, it investigates the importance of market structure in the sector that
does investment on investment incentives of the firm. It highlights the importance of
competitive pressure from another source- the buying industry. So it is not just the
competitive environment immediately around the firm that affects investment incen-
tives. A broader definition of competitiveness is required to fully assess the response
of firms to domestic and foreign competition. Secondly, the strategic complemen-
tarity of foreign and domestic competition is an important theoretical result that
points towards the importance of realistic modeling of the institutional environment
in which firms operate.
Thirdly, this dissertation is one of the first empirical investigations into the
microeconomic impact of industrial licensing (and hence de-licensing) on produc-
tivity of Indian firms. Given the scope of the licensing regime in India in terms
of the number of firms it impacted in a number of ways as well as in terms of the
time span of the ?License Raj?, it is, a priori, an interesting unexplored question.
Further, we use two unique data sets to provide conclusive evidence of the link be-
tween industrial deregulation and higher productivity in Indian manufacturing and
the strategic complementarity relationship between industrial and trade policies.
6
There are several questions that remain unanswered about the Indian policy experi-
ments. However, this dissertation is an initial attempt to analyze and quantify some
important facts about India.
7
Chapter 2
Strategic Complementarities between Trade and Industrial Policies:
A Theoretical Investigation
This chapter investigates theoretically the link between industrial policies that affect
the degree of competition in an industry and trade reforms with respect to their
impact on firm-level incentives to invest in quality-enhancing technology. We try
to capture significant features of an economy where industrial regulation takes the
form of restrictions on the entry of new firms.
Following Aghion et al. (2003b), we set up a two sector growth model. The
intermediate goods sector invests in quality-enhancing technological progress and
is assumed to be monopolistically competitive. The monopoly power of each pro-
ducer in the intermediate goods sector provides this producer with incentive to
invest in quality-enhancing technology. In the benchmark case, the final goods sec-
tor (that buys intermediate goods varieties for production) is modeled as perfectly
competitive. However in the presence of industrial policy that regulates entry into
manufacturing industry (for example, the industrial licensing regime in India) this
assumption is not likely to be valid in most cases. So we model the final goods
sector as a monopoly.
We find that our model gives substantially different results than the benchmark
case. A rise in competition in the intermediate goods sector (that invests in quality-
8
enhancing technological progress) will, under certain conditions, lead to a rise in
investment. This result is different from the standard Schumpeterian hypothesis that
a rise in competition should, unambiguously, reduce firm-level incentives to invest
in technology since the ability of the firm to appropriate surplus from investment
falls as competition rises. Given that the final goods sector has market power and
can control its output level, the intermediate goods firm faces two contradictory
forces. On the one hand, increased competition reduces the mark-up over costs
that it can charge and hence, reduces the returns from investment. However, as the
price-cost margin in the intermediate goods sector gets squeezed, derived demand
for its product from the final goods sector rises. This in turn raises the incentive to
invest in technology.
The second important result is that a rise in competition in the final goods
sector also leads to a rise in the investment incentives of the intermediate goods
firm. Since the output of the final goods sector now plays an important role in
determining the returns to investment, a rise in output in the final goods sector
(due to a rise in domestic competition) encourages the intermediate goods firm to
invest in technology.
Finally, we findthat undercertain conditions, trade liberalization and domestic
competition are strategic complements. That is, a rise in domestic competition
raises the marginal investment response of the intermediate goods firm to a rise in
the threat of foreign entry. The intuition behind this result is that more domestic
competition forces the firm to make productivity or quality decisions that in turn
help it respond to high quality foreign competitors.
9
An important point to note is that the framework presented in this paper
is applicable to a wide variety of distortionary domestic policies. In Chapter 1
we mentioned the specific case of India and its policy of industrial licensing that
affected market structure and incentives as a motivation for this dissertation. How-
ever policies like regulation of industry, subsidies to domestic producers, research
and development support to incumbent producers etc, which distort market struc-
ture and affect firm incentives to invest and innovate, can all be analyzed within
our framework.
2.1 Literature
There has been a recent stream of literature in industrial organization and endoge-
nous growth examining the relationship between product market competition and
innovation. Aghion et al. (2001) analyze the interplay between innovation and prod-
uct market competition and finds that product market competition (that is, how
substitutable two goods are in the consumers demand function) enhances innovation
in sectors where firms were already close to the technological frontier and discour-
ages innovation in sectors where firms are below the frontier. Aghion et al. (2003a)
use a model of step-by-step innovation to finds an inverted-U shaped relationship
between product market competition and innovation. This is supported by data on
firms in the UK.
While these papers provide an interesting framework in which to analyze the
issue of competition and innovation, they do not address the question of whether
10
their results are eroded or enhanced as a result of competition from high technology
foreign firms. That is, what is the interaction between domestic competition and
foreign competition? To our knowledge ours is one of the few papers that investigates
the issue of the interaction between foreign and domestic competition via their
impact on innovation incentives. Further, these papers do not investigate the issue of
interactions between the market structures in upstream and downstream industries.
Aghion et al. (2003b) and Aghion et al. (2004) are of particular relevance to
our results. Using the theoretical framework of Acemoglu et al. (2003), they assess
the impact of the institution of minimum wage laws on firm-level investment and
use the 1991 trade liberalization in India to illustrate how reform may have unequal
effects on industries and regions. Their main theoretical results are that liberaliza-
tion enhances investment in industries that were initially close to the technological
frontier and that pro-worker legislation lowers investment and this negative effect is
magnified by liberalization. Their empirical results confirm the main predictions of
the model. However, as mentioned earlier their theoretical model does not try to
model industrial policy in India and how this affected market structure and hence
innovation incentives. The policy of close control of private sector initiative was in
place for nearly 40 years and affected all aspects of production, distribution and even
consumption. Our model tries to capture the effect of policy on market structure
and thus attempts to provide a richer characterization of innovation incentives.
Melitz (2003) shows how a trade liberalization leads to reallocation of resources
across heterogenous firms. High productivity firms invest more as a result of liber-
alization and try to enter foreign markets while low productivity firms exit. Thus
11
similar to our model, high productivity firms have a better chance to survive in a
more competitive environment. Our model however, investigates whether the initial
state of ?high? productivity was a result of domestic industrial policy rather than an
exogenous event. This is particularly relevant for the case of India where monopolies
and oligopolies were created by industrial policy.
2.2 Model
The model builds on Acemoglu et al. (2003), Aghion et al. (2003b) and Aghion
et al. (2004). It is a discrete time growth model where foreign firms are allowed
to enter each period and sell their goods. Briefly, there are two sectors in the
economy. The intermediate goods sector (henceforth, IG) produces varieties of the
intermediate good (that differ in quality) and these are used by the final goods
sector (henceforth, FG) in production. Note that the FG producer cares about the
quantity and quality of the intermediate goods.
In the IG sector there is a monopolist in each variety of the intermediate good
and this monopolist chooses inputs (capital and labor) to maximize profits at the
beginning of each period, keeping in mind the derived demand of its product from
the final goods sector.
Within the same period, this monopolist also chooses the amount of investment
in quality-enhancing technology to maximize expected profits. The expectation is
over the probability of entry of foreign varieties that could replace the domestic
monopolist. The foreign firm decides whether or not to enter after observing the
12
investment in quality by the domestic intermediate producer but within the same
time period.
It is important to note that there is no profit maximization across time periods
in this model. That is, the entire maximization problem described above is repeated
each period. However, past investment decisions do affect IG firms since last period?s
quality determines the level of quality that the firm can achieve in this period. This
is because technology evolves in a discrete manner. The details of the model are
described in sections below.
2.2.1 Finals Goods Sector
All agents live for one period. There are two sectors in the economy. Following
Acemoglu et al. (2003), there is a unique final good in the economy and the final
goods (henceforth, FG) sector uses labor and a continuum of intermediate goods to
produce output according to the aggregate production function:
yt = L0t1??
bracketleftbiggintegraldisplay 1
0
At(v)?xt(v)?dv
bracketrightbigg
(2.1)
Here L0t is the number of production workers in the final goods sector at time
t. We assume that there is a fixed supply of labor in the economy ?L that must be
split between the FG and IG sectors. xt(v) is the quantity of intermediate input
produced in sector v and date t. At(v) is the quality of the intermediate input v in
producing the final good and ? and ? belong to (0,1). The final good can be used
either for consumption or as an input in the production of intermediate goods or for
investments in innovations.
13
An important feature of the production function is that it allows for the pos-
sibility of increasing, decreasing or constant returns to scale to the composite inter-
mediate good. That is, ? +? can be greater than, equal to or less than unity.
The next three sections describe the intermediate goods sector, the process of
technological innovation and the process of entry by foreign competitors. These are
derived from Aghion et al. (2003a).
2.2.2 Intermediate Goods Sector
Following Acemoglu et al. (2003) and Aghion et al. (2003b), in each intermediate
good sector v there exists a monopolist firm. There are many different varieties v of
the intermediate good and each variety differs from the other in terms of its quality
A(v). That is, the intermediate goods sector is monopolistically competitive. This
is because the FG sector demands an IG variety based on its price and quality. So
the variable v refers both to an intermediate sector industry and to the intermediate
firm that is active in that sector. For simplicity, each variety v can have one of two
qualities?advanced or backward. We explain this more in the next section.
As mentioned earlier, intermediate producers live for one period only and there
is no profit maximization across periods. The production technology uses labor and
capital to produce xt(v) units of the good v:
xt(v) = kt(v)?lt(v)1?? (2.2)
In each period, the intermediate goods producer in each variety observes the
derived demand for its product from the final goods sector and chooses its optimal
14
inputs of capital and labor. Then within the same period, it observes the threat of
potential entry from foreign firms and maximizes its expected profits to chose the
optimal level of investment in technology. The section below describes the process
of investment in greater detail.
An important point to note with regard to the assumption of monopolistic
competition in the IG sector is that this is a feature of Schumpeterian models. The
basic point is that some market power is required for any firm to have incentive to
invest in technology and this in turn, generates the standard result that a decline
in market power reduces the incentives of firms to invest.
2.2.3 Investment by Intermediate Goods Sector:
Production of Quality
As mentioned earlier, the numerous producers of intermediate goods compete in
quality. To simplify the model, IG firms differ in quality by discrete amounts.
That is, in each period domestic IG firms differ in their current distance from the
?technological frontier??defined as the highest quality under which the variety is
produced anywhere in the world. Note that foreign firms are assumed to be on the
frontier and thus, have the highest quality goods. The productivity of the frontier
technology in period t is given by ?At. We assume that this frontier grows at the
exogenous rate g. That is
?At = (1 +g) ?At?1 (2.3)
15
Here g is the rate of growth of rate of global technological advance. Further we
assume that a producer of any variety v of the IG can be on the frontier or just one
step below it. That is, at the beginning of period t, the IG firm v can be in any of
two states:
? ?Advanced? with productivity At?1(v) = ?At?1. These are the firms on the
current technological frontier.
? ?Backward? with productivity At?1(v) = ?At?2. These are firms that are one
step behind the frontier.
Before they decide production levels for period t, the firms must decide whether
or not to undertake innovative investments to enhance their productivity. The in-
vestments have stochastic returns where the probability of success equals the invest-
ment intensity zt. The cost function for investment is quadratic in the investment
intensity and is a function of the current level of technology.
ct(v) = 12zt2 ?At?1(v) ?1?? (2.4)
If research is successful then the incumbent firm can adopt the next most
productive technology. Thus a firm that was on the frontier in period t-1 continues
to be a frontier firm while a firm that was backward in t-1 advances to the level of
the advanced firm in t-1. If investment is not successful then the firm continues to
produce at the initial productivity.
Following Aghion et al. (2003b), we make the following simplifying assump-
tions about firm dynamics.
16
? If an advanced firm is successful at time t-1 then it starts as an advanced firm
at time t. All other firms start as backward firms.
? Note the implicit assumption of spillovers because firms that were backward
at time t produce at productivity ?At?2 in time t+1 rather than at ?At?3.
The following matrix represents the four possibilities that a particular IG variety
firm can be in period t, conditional on its productivity at the end of period t?1.
Initial Productivity Advanced in t?1 Backward in t?1
Invest successfully in t Initial= ?At?1 Initial= ?At?2
Final=(1 +g) ?At?1 = ?At Final=(1 +g) ?At?2 = ?At?1
Invest unsuccessfully in t Initial= ?At?1 Initial= ?At?2
Final= ?At Final= ?At?1 (spill-overs)
Table 2.1: Productivity of IG Firm
2.2.4 Foreign Entry into the Intermediate Goods Sector
We assume that entry by a foreign firm is such that the foreign firm does not per-
manently replace local producers. It enters to sell its product (produced elsewhere)
and leaves. This means that the foreign firm will not directly affect the distribu-
tion of productivity in the next period. One interpretation of this product entry
assumption is that entry threat is primarily due to lower barriers to trade in the
economy. Another possibility is that product differentiation and knowledge of local
market conditions allows domestic producers to remain in the market after foreign
entry.
17
Foreign firms in each variety of the intermediate good observe the outcome
of investment by domestic firms in period t before deciding whether to stay out or
to pay an entry fee ? and enter and be allowed to sell with probability ? in the
same period. This ? can be thought of as ease of entry into the domestic market.
A higher ? denotes lower tariffs, higher import quotas and any other regulations
which facilitate foreign entry and is consistent with our application to the case of
India which followed highly restrictive trade policies up to very recently. For the
purpose of this preliminary investigation, ? is exogenous and equal across sectors.
In particular it is assumed to be the same for sectors with a backward domestic
producer and sectors with an advanced domestic producer.
Further, the foreign firm is assumed to have the highest quality product. That
is, he operates on the technology frontier with productivity or quality ?At1. The setup
of the model then implies that if the foreign firm decides to enter the market, he
can displace the domestic producer of that variety, depending on the quality of the
domestic producer and the entry cost ?. If the foreign firm (FF) enters and competes
with a backward firm (that has quality lower than the technological frontier) then
FF gains the entire market. If FF enters and competes with an Advanced firm (that
is on the technological frontier and hence has the same quality as the foreign firm)
then Bertrand competition implies that the profits of both firms are zero.
Following Aghion et al. (2003b), we assume that the value of parameters is
such that the FF will always enter if the domestic firm (DF) is backward (that is, ?
is sufficiently small) and FF will never enter if DF is advanced and on the frontier.
1Note that the frontier itself advances at a rate of g (Equation 2.3).
18
Thus, the end of the period probability of entry into the market for v is
Prob(Entry in period t) =
??
???
????
0 if DF is advanced in t?1 and invests successfully in t
? otherwise
These probabilities will be used by the IG firm in order to maximize its expected
profits.
2.3 Market Structure
Our central goal is to analyze whether the incentives of IG firms to invest are
invariant to the particular market structure that is assumed for their industry and
for the FG industry. The price of the intermediate goods variety v, i.e., p(v) will be
a measure of the threat of potential entry from domestic competitors into the IG
sector while the parameter ? will measure the threat of potential entry from foreign
competitors. Intuitively, each firm in the IG industry will decide how much to invest
depending on how much of the gains from investment it can appropriate. These gain
depend on p(v)? the price that the firm can charge. This price, in turn, depends
on the market structure in that industry. Further, investment will also depend on
the market structure in the FG industry because market structure may determine
implicit bargaining power in the negotiations betweens the two sectors regarding
the split of surplus from investment. In this model, we use the equilibrium output
of the FG sector, y(.) as an indicator of market structure in the FG industry.
In the sections below we first solve the intermediate goods monopolist?s prob-
lem assuming that the final goods sector is perfectly competitive and then we solve
19
the model for the case where the final goods sector is monopolist.
2.3.1 The Benchmark Case: Competitive FG, Monopolist IG
Following Aghion et al. (2003b), we first derive equilibrium investment in technology
for the case where the FG sector is perfectly competitive. In all the analysis below
we assume that wages are the same for both sectors-intermediate and final. That
is, there is perfect mobility of labor between the two sectors. Further, in this case
we assume that the final good is the numeraire so py = 1. Perfect competition in
the FG sector implies that payment to inputs must equal the marginal product of
inputs. Using the production function of the FG sector (Equation 2.1), the price of
the intermediate good v is
pt(v) = ?LOt1?? At(v)
?
xt(v)1?? (2.5)
Similarly for labor,
wt = (1??)LOt??
bracketleftbiggintegraldisplay 1
0
At(v)?xt(v)?dv
bracketrightbigg
(2.6)
We can use Equation 2.5 to get the derived demand by the FG sector for intermediate
good v
xt(v) = LOt
bracketleftBigg?A
t(v)
?
pt(v)
bracketrightBigg 11??
(2.7)
An important point is that the elasticity of demand for variety v with respect to its
price is constant. That is, ? = 11??.
We assume that the IG firm is a price taker in the market for labor and
capital and that he chooses his inputs lt(v) and kt(v) to maximize profits. The
20
profit function of IG firm v can then be written as
?t(v) = Maxkt(v),lt(v){pt(v)kt(v)?lt(v)1?? ?kt(v)?wtlt(v)} (2.8)
s.t. kt(v)?lt(v)1?? ? LOt
bracketleftBigg?A
t(v)
?
pt(v)
bracketrightBigg 11??
From the first order conditions for profit maximization, we find that the derived
demands for labor and capital by the IG monopolist are given by lt(v) =
bracketleftBig1??
wt?
bracketrightBig?
xt(v)
and kt(v) =
bracketleftBig1??
wt?
bracketrightBig??1
xt(v) where xt(v) is the demand for the variety v by the final
goods sector. Putting these into the profit function we get
?t(v) =
?
?pt(v)?
parenleftBigg w?
1??
parenrightBigg??
?
parenleftBigg w?
1??
parenrightBigg1???
?xt(v) (2.9)
Thus the intermediate goods monopolist has constant marginal cost of production
? ?
parenleftBig w?
1??
parenrightBig??
+
parenleftBig w?
1??
parenrightBig1??
. Further, the each monopolist faces a derived demand curve
that has constant price elasticity. Thus profit maximization will lead the monopolist
to choose a constant mark-up over marginal costs. That is, pt(v)??pt(v) ? ? = 11??.
This leads us to an important result that the price of each and every variety of
intermediate good is equal.
pt(v) = 1?? ? ? for all varieties v (2.10)
This parameter ? is a measure of the extent of domestic competition that an inter-
mediate goods producer faces. A decline in ? can be thought of as a decline in the
price-cost margin that the intermediate firm can charge. Another possibility is that
? is the highest price that the domestic monopolist can charge without inducing en-
try by other domestic firms2. From Equation 2.1 we note that ? is a measure of the
2For example, if there is a competitive fringe in the intermediate goods sector that is less efficient
21
degree of substitutability between varieties of the IG in the final goods production
function. Thus, a rise in substitutability reduces the price that each producer v can
charge.
Before going further, we can solve for the derived demand for labor in the FG
goods sector (L0) by using the full employment condition for labor. From the profit
maximization exercise of the IG producer, the demand for labor by the producer of
variety v is lt(v) =
bracketleftBig1??
?wt
bracketrightBig?
xt(v)d. So the total labor demand in the IG sector is
LIG =
integraldisplay 1
0
(lt(v))dv =
bracketleftBigg1??
?wt
bracketrightBigg?integraldisplay 1
0
LOt
bracketleftBigg?A
t(v)
?
pt(v)
bracketrightBigg 11??
= ? 11??
bracketleftBigg1??
?wt
bracketrightBigg?
? ?11??At?Lot
The labor demand for the FG sector can be found using the full employment con-
dition
L0 +LIG = ?L (2.11)
Thus the demand for labor in the FG sector is given by the expression below.
LO =
?
?1 +? 11??
bracketleftBigg1??
?wt
bracketrightBigg?
? ?11??At
?
?
?1
?L (2.12)
The equilibrium production of the final good is given by
yt = ?
1
1??
? ?
??
1??LotA?t (2.13)
where
A?t ?
integraldisplay 1
0
At(v) ?1??dv (2.14)
in production and hence makes negative profits at the current price. Then ? will be the maximum
price the incumbent producer can charge while still keeping out the competitive fringe.
22
is the average productivity in the IG industry.
Now, after substituting for pt(v) = ? into the profit function and the derived
demand function for variety v, we can write profits of the IG firm as
?t(v) = ? 11?? (???)? ?11??LOtAt(v) ?1?? (2.15)
Equilibrium Investment
In Section 2.2.4, we described the rules that govern the entry of foreign firms into
the market. If the foreign firm is competing with an advanced, high quality domestic
firm who invested in technology and succeeded in maintaining its position at the
global technological frontier, then we assume that there is Bertrand competition
between the foreign and domestic firms and that the foreign firm will not enter at
all in those cases (given an entry cost ?). However when faced with a backward
domestic adversary, the foreign firm will enter and take over the entire market in
that particular variety. Further note that the probability that a domestic firm v will
innovate successfully, is equal to the intensity of investment, z. This means that the
only cases when a backward firm earns non-zero profits in period t are
? If it invests successfully in period t (and achieves quality At = ?At?1)?this
happens with probability z? and foreign firms stay out?probability (1 ? ?).
Thus the probability that a backward firm invests successfully and the foreign
firm does not enter is z(1 ? ?). In this case the domestic monopolist earns
payoffs ?t(v) = ? 11?? (???)? ?11??LOt ?A
?
1??
t?1 .
? If it invests unsuccessfully in period t (with quality At = ?At?2) and foreign
23
firms stay out. This happens with probability (1?z)(1??) and the payoffs
are ?t(v) = ? 11?? (???)? ?11??LOt ?A
?
1??
t?2 .
In the other two cases where foreign firms do enter, the incumbent backward firm
earns zero profits.
Similarly, the cases when an advanced firm earns non-zero profits are
? If it invests successfully in period t and foreign firms stay out. This hap-
pens with probability z(1 ? ?). However we assume that the values of pa-
rameters are such that the foreign firms will never enter if domestic firm is
on the frontier. That is, ? = 0. The domestic firm earns payoffs ?t(v) =
? 11?? (???)? ?11??LOt ?A
?
1??
t .
? If it invests unsuccessfully in period t and foreign firms stay out. This happens
with probability (1?z)(1??) and the payoffs are
?t(v) = ? 11?? (???)? ?11??LOt ?A
?
1??
t?1 .
A firm chooses investment zt to maximize its expected profits. The last term
in the equation below represents the costs of investing zt (Equation 2.4).
E?t(v,Backward) = ? 11?? (???)? ?11??LOt
bracketleftbigg
z(1??) ?A
?
1??
t?1 + (1?z)(1??) ?A
?
1??
t?2
bracketrightbigg
?12z2 ?A
?
1??
t?2 (2.16)
This gives equilibrium intensity as
zB = ? 11?? (???)? ?11??LOt(1??)(1 +g) ?1?? (2.17)
Note we use the fact that ?At?1 = (1 +g) ?At?2.
24
Similarly the expected profit for an advanced firm in period t is
E?t(v,Advanced) = ? 11?? (???)? ?11??LOt
bracketleftbigg
z ?A
?
1??
t?1 + (1?z)(1??) ?A
?
1??
t?2
bracketrightbigg
?12z2 ?A
?
1??
t?2 (2.18)
This gives equilibrium intensity as
zA = ? 11?? (???)? ?11??LOt
bracketleftBig
(1 +g) ?1?? ?(1??)
bracketrightBig
(2.19)
Equations 2.19 and 2.17 can be used to analyze the effects of competition on
investment in productivity-enhancing technology 3.
Comparative Statics
Analyzing the expressions for equilibrium investment in quality-enhancing technol-
ogy we find that similar to Aghion et al. (2003b), there is differential response to
a trade liberalization episode across firms based on the initial level of quality that
they start from. That is, ?zA?? > 0 and ?zB?? < 0. Under increased threat of entry,
advanced firms raise their investment since they know that there exists a chance
that their investment will be successful and the foreign firm will not enter. On the
other hand, backwards firm reduce investment when they perceive higher threat
of entry. This is because their expected gains from the investment are lower be-
cause there is a higher chance that the FF will enter and they will earn zero profits.
3Note that the market clearing wage rate can be calculated using Equation 2.6 as wt =
1??
? ?
??
1??A?t where A?t ?
integraltext1
0 At(v)
?
1??dv is the average quality in the intermediate goods industry
at time t.
25
Thus post-liberalization, productivity should be higher in industries with higher
pre-liberalization productivity.
Further, the standard Schumpeterian motive for investment holds- the higher
the price that the intermediate goods producer can charge for his product, the larger
the share of surplus that he can appropriate and hence, the higher the returns to
investment in quality. That is,
dzA
dw > 0 ?
dzA
d? > 0 if ? > (1?(1??)
2)? ? ? ? ?? (2.20)
The details of this derivation are presented in Appendix A.1.
Moreover, given the previous result it is easily seen that domestic competition
as measured by ? and foreign competition as measured by ? are strategic substitutes.
That is,
d2zi
d?d?
??
???
????
> 0 for i=Advanced
< 0 for i=Backward
(2.21)
The more monopoly power in the IG sector or the higher the threat of potential
entry into the IG sector (as measured by a higher ?), the more aggressive will be
its response to the increased threat of foreign entry (measured by a higher ?).
2.3.2 Monopoly in the FG Sector
We investigate the implication of assuming that the FG sector, that buys interme-
diate goods, has some market power. For this preliminary theoretical exercise, we
assume that the FG sector is monopolist in the market for the final good. That
is, the FG producer faces a downward sloping demand for his good. The demand
26
function is linear 4. We discuss the choice of FG output by the monopolist below.
The output of the final goods sector?y(.)?is an important variable in our analysis.
In order to minimize costs, the FG producer chooses xt(v) (taking the price of
inputs w and p(v) as given5 to solve
C(p(v),w,y) = MinLot,x(v),v?(0,1)wtLot +
integraldisplay 1
0
p(v)x(v)dv (2.22)
subject to producing enough so that the FG monopolist can produce his monopoly
quantity y. That is,
yt ? L0t1??
bracketleftbiggintegraldisplay 1
0
At(v)?xt(v)?dv
bracketrightbigg
(2.23)
The details of the cost minimization are in Appendix A.2. Using the first order
conditions and the constraints, we are able to derive the expression for the derived
demand for variety v as
xt(v) = ?wLot1??
?
?At(v)
?
1??
pt(v) 11??
?
?
?
?
integraldisplay 1
0
At(v) ?1??
pt(v) ?1??
?
?
?1
The first thing to note about the derived demand function is that given the property
that a monopolistically competitive IG producer does not take account of the effect
of his actions on the average price (Tirole (1988)), the elasticity of demand with
respect to price is constant and equal to 11??. Thus, a profit maximizing IG producer
who faces constant marginal cost of production and a constant elasticity of demand
4The demand curve takes the form Py = a?by.
5We assume that the monopolist in the FG sector takes p(v) as given. That is, the FG firm does
not use its monopsony power over the IG sector. Though this is an interesting case to think about,
it may be algebraically untractable as is the case with models with many layers of competition.
27
will charge a constant mark-up over costs. Thus exactly similar to the perfectly
competitive FG case,
pt(v) = ? = 1?? for all varieties v (2.24)
Substituting this into the derived demand equation and using the FG produc-
tion function gives us
xt(v) = ?wLot1??
?
?At(v)
?
1??
?A?
?
? =
bracketleftbigg w?
1??
bracketrightbigg1???
?A(v)
?
1??ym
?1??A?2??
?
? (2.25)
where ym is the equilibrium monopoly output for the final goods sector. The cost
function of the final goods producer can then be written as
C(p(v),y) =
bracketleftBigg
w
parenleftbigg1??
w?
parenrightbigg?
+
parenleftbigg1??
w?
parenrightbigg??1bracketrightBigg
??A???1ym (2.26)
Here ? is given by Equation 2.24. Thus the final goods monopolist has a constant
marginal cost of production. Note that the total cost rises in ?- the price of the
inputs and falls in A?- the average quality or productivity of the inputs.
Now profit maximization for the final goods monopolist implies that he will
choose ym to equate marginal revenue to marginal cost 6. That is,
a?2bym =
bracketleftBigg
w
parenleftbigg1??
w?
parenrightbigg?
+
parenleftbigg1??
w?
parenrightbigg??1bracketrightBigg
??A???1
Thus, equilibrium final goods output is given by
ym = 12b
bracketleftbigg
a? M?
?
A?1??
bracketrightbigg
(2.27)
6The demand curve facing the final goods monopolist is Py = a?by. So marginal revenue is
a?2by.
28
where M ? w
parenleftBig1??
w?
parenrightBig?
+
parenleftBig1??
w?
parenrightBig??1
. Equations 2.25 and 2.27 give us a full description
of the final goods sectors demand for intermediate goods.
Given that the derived demand for each intermediate goods variety depends
on the output of the final goods sector which in turn depends on the price and the
quality of the variety, the level of output ym can be used as a measure of the market
power of the final goods monopolist in its relationship with the intermediate goods
sector. The intuition for this result is that higher price or lower quality of a variety
will reduce the optimal output of the FG sector and will in turn exert pressure on
the IG sector to either reduce price or increase quality by investing more. Thus,
its scale of operation is the bargaining chip that the FG sector uses against the IG
sector producers.
Equilibrium Investment
The intermediate goods producers profit maximizing exercise is similar to the case
where the final goods producer is perfectly competitive. His profit function after
optimal choice of labor and capital has been made is
?t(v) = [???]xt(v)
where ? ?
parenleftBig w?
1??
parenrightBig??
+
parenleftBig w?
1??
parenrightBig1??
is the constant marginal cost of production. Other
than the derived demand for the intermediate good, nothing has changed for the
intermediate goods producer. Following the same procedure as in Section 2.3.1, the
equilibrium investment in productivity-enhancing technology by the intermediate
29
goods producer is given by
zB = (???)???1
parenleftbigg w?
1??
parenrightbigg1?? (1??)(1 +g) ?1??
A?2?? ym (2.28)
zA = (???)???1
parenleftbigg w?
1??
parenrightbigg1?? [(1 +g) ?1?? ?(1??)]
A?2?? ym (2.29)
2.3.3 Results
In this section we analyze the impact of market structure in the final goods sector
on equilibrium investment in the intermediate goods sector.
Foreign Entry
In the model, the probability that a foreign firm enters the domestic market to sell
its product is given by ?. As mentioned earlier, ? is an exogenous parameter and
can be thought of as a measure of trade openness. Thus, an increase in ? could be
due to a trade liberalization episode. From Equations 2.28 and 2.29, we see that
the basic effect of the threat of foreign entry is the same as in the original model
(Aghion et al. (2003b)). Advanced firms are spurred by the threat of entry to invest
more in order to deter foreign entry while backward firms invest less because there
is a lower chance of them being able to deter entry and hence lower expected payoffs
to investment.
30
Effect of Domestic Competition
An important parameter in our model is ?, the price that the IG producer charges
for the variety that he produces. As we show in Section 2.3.2 (Equation 2.25) and
Appendix A.2, the derived demand for each variety is such that the IG monopolists
in all varieties charge a common price ?. Now we conduct an interesting thought
experiment and allow ? to change.
As mentioned earlier, ? can be interpreted as a measure of the degree of
domestic competition in the IG sector. Suppose that the IG monopolist is charging
a constant mark-up over marginal cost. If entry occurs into the IG sector, it is
plausible that this mark-up falls (either due to rise in costs or decline in prices)7.
Another way that increased competition may affect ? is in the long term.
As entry in to the intermediate goods sector continues, the elasticity of demand
for each intermediate good variety may rise. That is, the market for intermediate
goods starts saturating. This in turn will lower the price-cost margin that each
intermediate goods producer can charge.
As shown by Equation 2.24, the price ? is a function of w. So in order to
7Note that there can not be entry in the model since the IG sector requires market power in
order to invest in quality-enhancing technology. However, a fall in ? mimics the impact of actual
entry on IG firms. Further, Aghion et al. (2003b) assume that the IG sector is monopolistically
competitive but is surrounded by a higher-cost, lower-quality ?competitive fringe?. In this set
up, ? can be interpreted as the highest price that the monopolist in variety v can charge without
inducing entry by producers in the competitive fringe. That is, at any price higher than ? even
the not-so-productive firms in the fringe can sell profitably and hence have incentive to enter.
31
totally differentiate equilibrium investment (Equations 2.28 and 2.29) with respect
to ? and to maintain algebraic tractability, we analyze a special case where a change
in ? occurs due to a changes in w and ?L (the endowment of labor in the economy)
that leave the allocation of labor between the IG and FG sectors unchanged. We
arrive at the following proposition8.
Proposition 1: Under the assumption that the FG sector is monopolistic and the
IG sector is monopolistically competitive, we find that
dzi
d? = C
parenleftbigg w?
1??
parenrightbigg1??
???1
bracketleftBigg(1??)?y
m
? +
(1??)(???)ym
w
dw
d? + (???)
dym
d?
bracketrightBigg
< 0 for ? > ??,i = Advanced,Backward
i.e., more competition in the IG sector (lower ?) leads to higher investment in
technology if the initial entry barriers are above a threshold.
The first term in the square brackets is the standard Schumpeterian effect-
lower price means that the IG monopolist is appropriating a smaller share of the
surplus and hence will invest less. This effect was present in the benchmark case
when the FG sector was assumed to be perfectly competitive. The second term
shows the direct impact of a change in wages (due to a change in labor endowment)
on investment in the IG sector. A rise in wages causes the final goods sector to
substitute out of labor into the intermediate good. This means a rise in the derived
demand for intermediates. This allows the IG producer to raise the price of inter-
mediates ?. Thus the direct effect of a change in wages that changes ? bolsters the
standard Schumpeterian result by providing greater incentive to the IG monopolist
8Details are presented in Appendix A.3
32
to invest in technology.
The third term is unique to the case where FG sector is monopolistic. This
is the Bi-lateral monopoly effect?a lower price of intermediates will raise the out-
put of the FG goods sector and raise the profits/investment of the IG producer.
The main intuition for this result is that while the perfectly competitive FG sector
takes account of the price and average quality of the intermediate goods ex post
(Equation 2.13 shows that total FG output decreases in the price and rises in aver-
age quality), monopoly power in the FG sector allows the producer to take ex ante
account of the price and average quality and chose its output accordingly. Thus,
the FG monopolist chooses its output level (Equation 2.27) and hence, its derived
demand for each variety of intermediates, taking into account the price of the variety
?, the quality of the variety A(v) and the average quality in the intermediate goods
sector A? and forces the IG producer to respond via investment in quality 9.
The bi-lateral monopoly effect outweighs the Schumpeterian effect when the
initial price distortion in the IG sector is very high (? > ??). If the initial price is
very high and if it falls even a little, there will be a large response of FG output y(.)
to this and the larger sale volume will compensate the IG producer for the decline
in price. This is borne out by the fact that the elasticity of y with respect to ? is
9Note that while we allow the FG producer to take advantage of his monopoly position in the
market for final goods, he does not use his monopsony power as the only buyer of intermediate
goods. While that would be an interesting case to examine, there is likely to be indeterminacy
in models with many layers of competition and we will not be able to get a simple, closed form
solution to the model.
33
strictly rising in ?10.
An interesting feature of this result is that it holds at the profit maximizing
price ? = 1??. That is, when the intermediate goods firm is charging its monopoly
price there is a positive level of final goods output ?y such that if final goods output is
below ?y then a fall in the price of the intermediate good will force the intermediate
goods firm to invest more in technology. This is because equilibrium output ym
(Equation 2.27) is strictly decreasing in ?. When we analyze the conditions when
ym ? ?y, we find that the average quality of intermediates A? must be lower than
some level. That is, ? ? ?? ? ym ? ?y ? A? ? ?A?. So a rise in domestic
competition forces the IG sector producer to increase his investment in productivity
when A? ? ?A?. Intuitively, the price of the intermediate goods matters to the final
goods firm. But, holding the price constant, the average quality of intermediate
goods also matters. Given that the intermediate producer is charging his monopoly
price, the average quality of his product needs to be higher than a threshold level
in order to induce adequate demand from the final goods buyer.
An important observation about the impact of competition can be made from
Equation 2.25. If we look at Equation 2.25 (that shows the derived demand for
variety v of the intermediates good) we notice that demand for variety v is directly
proportional to the quality of variety v (A(v)), and inversely proportional to the
index of average quality in the intermediate sector (A?). That is, the final goods
monopolist weighs the quality of each input against the average quality of inputs
that is available to him in the market. Holding ? constant, the demand for v will rise
10From Equation 2.27, the elasticity of y with respect to ? is given by ?y = ?M??
2bA?1??ym.
34
only if A(v) rises relative to A?. That is, the average quality in the intermediates
goods sector can be used as an index of competitive pressure in the intermediate
goods sector.
It is interesting to contrast this result with the one obtained when the final
goods sector was modeled as perfectly competitive. In Section 2.3.1 we found that
dzi
d? > 0 for all values of ?. That is, when the buying industry (final goods sector)
does not have any market power, only the standard Schumpeterian motive operates
and more monopoly power to the intermediates sector (measured by a higher ?)
will lead to higher productivity/quality in that sector. However when the buying
industry has some market power, it takes ex ante account of the price and quality of
each variety while determining its output level (ym) and hence, the derived demand
for the variety. This, in turn, allows the FG sector to get higher quality or lower
price products from the intermediate goods sector. Thus there exists a range where
even a fall in ? will spur investment in the intermediates sector11.
11In Appendix A.5 we investigate the case of a regulated monopolist in the final goods sector.
We assume that the government applies licensing to the final goods sector by constraining the FG
monopolist to produce on or below a certain level. We find that the supply curve of the monopolist
is perfectly inelastic (at the level of the government-given output) for cost (and hence ?, the price
of IG varieties) less than a certain level. For cost (and ?) above this level, the supply curve is the
same as the function derived in Section 2.3.2. Thus, for ? < ?C, a decline in market power in the
IG sector will lead to lower investment in technology in the IG sector. But for ? ? ?C, a fall in
? (maybe due to removal of entry restrictions in the IG sector) might have a positive impact on
investment in productivity. That is, with a more distorted IG sector (measured by a higher ?),
there is a higher chance of more competition fueling a rise in investment and productivity in the
35
If we analyze the equations for the derived demand for intermediate variety v in
the two market structures (Equations 2.7 and 2.25) we find an important difference.
The perfectly competitive firm does not take ex ante account of the average quality of
intermediate goods when it decides to buy variety v even though ex post, its output
depends on average quality (Equation 2.13). On the other hand, a monopolist firm
takes the average quality into account ex ante. More generally, we can think of A?
as price weighted average quality.
Interaction of Domestic Competition and Foreign Entry
We can analyze how a unit rise in domestic competition in the IG industry will
impact the responsiveness of IG firms to threat of foreign entry. We find that
Proposition 2: Under the assumption that the FG sector is monopolistic and the
IG sector is monopolistically competitive, we find that
d2zi
d?d?
??
???
????
< 0 for ? > ??,i = Advanced
> 0 for ? > ??,i = Backward
(2.30)
That is, the firm is more responsive to foreign entry when the domestic environment
is more competitive.
The derivation of this result is given in Appendix A.2. This is our main theoretical
result and this points towards a strategic complementarity relationship between
industrial and trade policy. That is, industrial deregulation (i.e., higher threat
of potential entry as measured by a decline in ?) raises the marginal investment
response of a firm to trade reform. The condition ? > ?? means more distortion in the
IG sector.
36
IG sector. Another way to think about it is that domestic and foreign competition
are strategic substitutes in the range ? < ?? and strategic complements in the range
? > ??.
The intuition behind this result is the balance between the standard Schum-
peterian hypothesis that a rise in price of the product raises incentives to invest and
the bilateral monopoly feature of the model that underlies the result in Proposition
1. A decline in price of the IG good reduces the incentives of the IG producer to
invest in quality. However, a lower price also raises derived demand for the product
and hence encourages investment.
Note that when the final goods sector was modeled as perfectly competitive,
we got the result that ? and ? are always strategic substitutes.
Effect of Market Structure in the Buying Industry
Another important variable in our model is ym(.)?the output of the final goods
sector producer. This provides an indirect measure of the market structure and
competition in the FG sector. Using Equations 2.28 and 2.29, we see that given
average quality in the intermediates good industry, a rise in the output of the up-
stream monopolist y leads to a rise in investment intensity of both types of firms.
Details are in Appendix A.4.
Proposition 3: Under the assumption that the FG sector is monopolistic and the
IG sector is monopolistically competitive, we find that
?zi
?ym(.) > 0, i = Advanced,Backward (2.31)
37
That is, the intermediate goods firm is affected not only by a change in competition
in its own industry, but also by a change in competition in its buying industry.
Thus, as the final goods industry moves from a monopoly to perfectly com-
petitive and final goods output rises, the investment incentives of the intermediate
goods sector rise, since derived demand for each variety rises with final goods sector
output.
Another Complementarity
In our model, we consider two possible measures of domestic competition. The first
one, ?, measures the level of competition in the intermediate goods sector. The
second one, ym, represents the degree of competition in the final goods sector. The
more the market power the final goods producer has, the lower will be his output
and hence the higher the price of his product py. As we have seen, a rise in output
of the final goods sector directly affects the derived demand for intermediates by
this sector and hence raises the pay-offs to investment in technology. Further we
find that,
Proposition 4: Under the assumption that the FG sector is monopolistic and the
IG sector is monopolistically competitive, we find that
?2zi
???ym
??
???
????
> 0 for i = Advanced
< 0 for i = Backward
(2.32)
That is, an intermediate goods firm is more responsive to foreign entry when he sells
to a more competitive final goods sector at home.
Thus, what matters to the intermediate goods producer is not only competi-
38
tion in the intermediate sector but also in the final goods sector. More market power
in the final goods sector (measured by lower ym) reduces the gains from investment
in technology (since some gains will now be appropriated by the final goods mo-
nopolist). As output in the final goods sector rises, the derived demand for each
variety of intermediates rises, raising incentives for investment. Now, when faced
with threat of foreign entry, the intermediate goods producer, who already has a
higher level of investment, will be spurred to invest more in technology in order to
deter entry by the foreign seller.
2.4 Robustness Checks
In this section we test how robust our theoretical results are to various assumptions
that we have made.
A particular feature of our model is that the generation of quality needs
investment?z?but no resources. That is, investment is similar to advertising ex-
penditure for the IG firm. We test the robustness of our results to a more general
specification of the quality-generation process and let production of quality use labor
so that we can check whether our results in previous sections are due to the fact that
investment is a manna from heaven. In that case, labor will need to be diverted
from the production of physical units of the intermediate good into research and
development in order to produce better quality/productivity. So we can think of
investment intensity z as the number of workers that were removed from production
and into research and development (or we can think of the production function for
39
quality as Probability of Successful innovation= z = lrd). The profit function (after
maximization with respect to capital and labor used in production of quantity) is
given by
?t(v) =
parenleftBigg ???
?1??A?t2??
parenrightBiggparenleftbigg w?
1??
1??parenrightbiggy
m ?wz ?
1
2z
2 ?A ?1??
t?1 (2.33)
Then maximization of expected utility with respect to z gives us equilibrium invest-
ment as
zA = (???)???1
parenleftbigg w?
1??
parenrightbigg1?? [(1 +g) ?1?? ?(1??)]
A?t2?? ym ?
w
?A ?1??t?1
(2.34)
Thus we see that equilibrium investment in the case where production of quality
requires labor is lower than investment in the case with no labor . The second term
on the right hand side reflects the marginal cost of each unit of z in terms of the cost
of workers that need to be hired12. All the results regarding domestic competition,
foreign competition and the relationship between the two remain essentially the
same as the ones described in Section 2.3.3. So the incentives to invest in quality in
response to domestic and foreign competition change magnitude but not signs.
12We can impose the full employment condition here to get the expression for L0 and hence for
y(.). The condition is that L0 +integraltext10 (li +zi)di = ?L where li is the equilibrium labor requirement for
production of quantity and zi is the labor required for R and D. Solving this equation for L0 we
find that output of the FG sector is lower than in the case where quality does not require labor.
This is because the IG sector now demands more labor (since it needs labor to produce quality
and quantity) and this cuts into the amount of labor that can be employed in the FG sector.
40
2.5 Conclusions
The main conclusion of this chapter is that assumptions about market structure in
all sectors of the economy matter. Under the assumption that the FG sector has
market power in the final goods market, we find that the investment incentives of the
intermediate goods producers are substantively different then the benchmark case.
The basic intuition underlying this is that control over output level matters. Derived
demand for intermediate goods depend on final good output. A monopolist final
goods firm uses the level of its output as a mechanism to induce the intermediate
goods firm to provide it better quality or lower price of inputs ex ante. When the
final goods sector is modeled as perfectly competitive, it takes only ex post account
of the price and quality of the goods that it buys and hence cannot use the bi-lateral
monopoly factor to get better quality from the intermediate goods sector.
Our study has several important implications. Firstly, seemingly innocuous
assumptions about market structure can matter. Secondly, competitive environment
needs to be defined more broadly to be able to have a richer characterization of
investment incentives. For example, our model shows that entry de-regulation in
the final goods sector can affect investment in the intermediate goods sector.
Further, we providearichercharacterizationoftheeffectsofentryde-regulation
in the intermediate goods sector. In the case where the final goods sector is per-
fectly competitive, any de-regulation in the intermediate goods sector that raises
competition (modeled as a fall in the price of intermediates) has the effect of un-
ambiguously lowering technological investment in the intermediate sector. How-
41
ever when we model the final goods sector to have ex ante control over its output
level by giving it some market power we find that under certain conditions, entry
de-regulation in the intermediate goods sector can raise the level of technological
investment. That is, under certain conditions it is possible to spur technological
development in the economy by reducing the market power of firms that invest in
technological development.
The last point to note is that the framework presented in this paper is appli-
cable to a wide variety of distortionary domestic policies. We mention the specific
case of India and its policy industrial licensing that affected market structure as a
motivation for this paper. However policies like regulation of industry, subsidies to
domestic producers, research and development support to incumbent producers etc,
which distort market structure and affect firm incentives to invest and innovate, can
all be analyzed within our framework.
42
Chapter 3
Competing or Collaborating Siblings? Industrial and Trade Policies
in India
3.1 Introduction
There has been intense debate over the past few decades about the impact of trade
liberalization on welfare, growth and productivity. This question is of immense
importance and relevance to developing countries, in particular large and populous
economies like China, India and Brazil that have recently entered the mainstream
of the world economy. Some argue that the contribution of trade towards raising
competition faced by firms is a way to promote growth in developing countries (for
example, Dollar and Kraay (2001), Frankel and Romer (1999), Sachs and Warner
(1995)). Others (Rodrik and Rodriguez (2001)) stress the importance of domestic
institutions and conditions in the success of trade liberalization.
This debate is further confounded by the debate about whether competition
itself enhances or erodes firm-level incentives to innovate and grow. The standard
argument that market power reduces firm-level incentive to innovate is pitted against
the Schumpeterian argument that the more the surplus a firm can appropriate (via
more market power) the greater the incentive to invest in technology. This is a
question of policy relevance not only for developing countries but also for developed
43
countries.
In this chapter, we are interested in the impact of domestic economic con-
ditions and of trade reforms on firm-level productivity for the case of India. The
specific economic condition that we investigate is the level of competition that a
particular firm or industry faces in the domestic market. Then we assess the impact
of competition from abroad (due to trade reforms). So the basic empirical questions
that this chapter tries to test are what is the first, most proximate impact of a rise in
competition as a result of industrial de-regulation that a firm faces on a commonly
used measure of productivity-output per worker and further, what is the relation-
ship between domestic and foreign competition? Can the benefits from trade reform
be enhanced by encouraging more competition within domestic industry?
As mentioned earlier, several large developing countries have recently under-
taken trade liberalization and deregulation. Our case study-India-is particularly
interesting and relevant for two reasons. As already mentioned, a very rigid and
stern industrial licensing regime was in operation for 40 years. Entry to into man-
ufacturing industry was not free (i.e. not based on market forces) and this in turn
lead to monopolistic distortions in almost all sectors. The second feature that makes
India a good case study for the relationship between trade and industrial reforms is
the interesting chronology of reforms- industrial de-regulation in the 1980s and trade
reforms in the 1990s. This chronology of reforms allows us to distinguish between
the two types of reforms and to assess the relationship between them.
It is important to note that though there have been several studies of the
impact of trade liberalization in India on firm-level productivity (e.g., Krishna and
44
Mitra (1996), P. Balakrishan and Babu (2000), Sivadasan (2003)), this is the first
attempt to assess the impact of the reforms of the 1980s. None of the previous
studies control for the changes in the 1980s and hence provide biased estimates
of the impact of trade reform. This issue is even more important in the light of
a puzzling empirical result. As Panagariya (2002), Panagariya (2004) and Delong
(2001) point out, reliable productivity measures show a sharp rise in productivity
levels and rates of growth prior to the trade reforms of the 1990s.
We use two unique data sets on India to test our predictions. The first data
set on industrial policy in India from 1970 onwards is to our knowledge, the only
comprehensive data set on industrial policy and one that allows econometric esti-
mation of the impact of deregulation. We are able to identify which industry was
deregulated in which year of the 1970s, 1980s and 1990s. Further we use a new
plant-level data set for the period 1980-1994. This is a census of firms in India and
has only recently been made available to researchers. The span of our data allows
us to capture the impact of the reforms of the 1980s and the 1990s.
The specific predictions we test are derived from theory. In Chapter 2 we use
a two sector Schumpeterian growth model based on Aghion et al. (2003b) and Ace-
moglu et al. (2003) and analyze firm-level incentives to invest in quality-enhancing
technology. The intermediate goods sector invests in technology and the final goods
sector production function includes both the quantity and the quality of these inter-
mediates. Proposition 1 states that given monopolistic distortion in the final goods
sector, more competition in the intermediate goods sector (modeled as a fall in the
price that the firm is able to charge for its product) can lead to a rise in technolog-
45
ical investment by the firm and hence to a rise in the productivity of the firm. In
the Indian case, firms in industries that were deregulated faced a more competitive,
market-based environment and these firms can be used to test our prediction.
Further, Proposition 2 gives us an indication of the impact of a rise in competi-
tion from foreign producers on the firm. Our second prediction is that under certain
conditions, industries that face more competition domestically (that is, industries
that were deregulated) tend to perform better in the face of foreign competition.
That is, there may be strategic complementarity between industrial de-regulation
and trade reform. The intuition behind this result is that competition at home
forces the firm to make investment in productivity-enhancing technology and that
these investments prepare the firm to compete with foreign entrants. The policy
implication of this prediction is that it may be possible to raise the benefits from a
trade liberalization episode by first facilitating greater competition domestically.
Our identification strategy to assess the impact of domestic deregulation on
firm-level productivity uses an important institutional feature of Indian policy.
Firms with assets below a certain defined rupee threshold were exempt from li-
censing requirements. Thus in any industry, some firms are ?non-licensed? or ?ex-
empt? (since they are below the licensing threshold). So in industries that were
?de-licensed? by the government, a firm is treated by de-licensing reform only if it
was large enough to be under licensing at the time of the reform. This institutional
feature provides within-industry variation that allows us to identify the interaction
between de-licensing and size. Given the nature of the licensing regime in India,
the coefficient on this interaction gives the joint impact of two mechanisms through
46
which de-licensing in India affected firm-level productivity. The first mechanism is
the removal of direct, microeconomic constraints on the firm (for example, output
limits) which impacted the ability of a firm to become more productive while the
second mechanism is the removal of entry restrictions on the industry as a whole
which impacted the incentives of firms to become productive.
Our main result is that piece-meal industrial de-regulation that took place in
the 1980s has a positive and significant impact on labor productivity. Further, pre-
liminary results suggest that there exists strategic complementarity between trade
liberalization in 1991 and industrial de-regulation. That is, industries that were
de-regulated tended to fare better after the trade liberalization episode.
We conduct a variety of robustness tests. Firstly, we test an implication of
our identification assumption that for the group of industries where exemption from
licensing was not granted, there should be no size-based response to de-licensing.
Secondly, we analyze the distribution of assets and the rates of growth of assets
of exempt firms over licensed and de-licensed industries and show that there was
no difference between the response of exempt firms in the two types of industries.
This means that the argument that firms choose their assets and hence, choose
their exemption status, based on anticipations of reform may not hold. We also
provide evidence that there was no bunching of a large number of firms under the
threshold and hence firms were not deliberately trying to stay exempt from licensing
requirements. Another test of our identification assumption is that there should be
no variation in the response of firms to de-licensing around arbitrary thresholds and
we find that to be the case in our data. Our results are robust to the inclusion of
47
a wide variety of firm-level controls (ownership, organization, location (urban/rural
and state) and wage to rental ratio) as well as industry-level controls (industry
fixed effects, entry for firms belonging to large houses etc) that may affect firm-level
productivity and/or be correlated with size. We also control for industry-level time-
varying factors (by including industry-year fixed effects). Since our data does not
allow us to follow a firm over time, we use certain firm-level identifiers to create a
pseudo-panel. We are able to identify several cross-sections of very similar firms in
our sample. Using the pseudo-panel we find that both our important results are
robust to the inclusion of firm-level fixed effects.
De-regulation can change investment decisions and productivity by many dif-
ferent mechanisms. The one implied by our theoretical model is an increase in the
threat of potential entry and higher competition from new firms as measured by
a reduction in the price-cost margin of the firm. Other mechanisms can include a
rise in capacity utilization if the firm had excess capacity prior to de-regulation, a
change in the capital to labor ratio in case regulation involved restrictions on the
level of capital a firm could use in production, a decline in corruption that releases
resources to be used for productive purposes etc. Though any full story about the
impact of a de-regulation episode should include details of the mechanism through
which de-regulation affects the firm, in this chapter we focus on the impact of de-
regulation on investment and productivity choices rather than on the mechanisms
through which this impact might occur.
48
3.2 Literature Survey
Our paper contributes to two main strands of literature. The first strand focuses
on the impact of competition -domestic or foreign- on firm-level productivity and
incentives. The second strand focuses on the impact of trade reforms on firm pro-
ductivity.
There has been a stream of recent papers that examine the relationship be-
tween competition and productivity. Heterogeneous firm models highlight how com-
petition could lead to higher aggregate productivity levels by forcing the exit of
unproductive firms and leading to reallocation of resources from less productive to
more productive firms (e.g., Hopenhayn (1992), Melitz (2003)). Representative firm
models highlight broadly two channels through which competition could affect firm
productivity: by providing better (worse) incentives for managers and workers to
reduce slack and cut costs e.g, Schmidt (1997) or by providing better (worse) incen-
tives for innovation or for adopting new technologies e.g., Aghion et al. (1999) and
Aghion and Howitt (1992). This strand of literature has examined the relationship
between product market competition and innovation. Aghion et al. (2001) analyze
the interplay between innovation and product market competition and finds that
product market competition (that is, how substitutable two goods are in the con-
sumers demand function) enhances innovation in sectors where firms were already
close to the technological frontier and discourages innovation in sectors where firms
are below the frontier. Aghion et al. (2003a) use a model of step-by-step innovation
to find an inverted-U shaped relationship between product market competition and
49
innovation. This is supported by data on firms in the UK. Nickell (1996) reviews
the studies that directly relate competition to productivity and finds a generally
positive effect of competition on productivity. Djankov and Murrell (2002) review
studies on the impact of product market competition on productivity in transition
economies and find that in general more competition improves firm performance.
Aghion et al. (2003b) use a Schumpeterian endogenous growth model to derive their
main theoretical predictions. This model consists of a perfectly competitive final
goods sector and a monopolistically competitive intermediate goods sector. They
find that industries that were closer to the technological frontier pre-reforms tend
to do better than other industries. Further, industries located in states that have
pro-business policies (for example, a lower minimum wage) tend to perform better.
Aghion et al. (2004) assess the impact of domestic deregulation on industry produc-
tivity for the case of India and find that industries that were deregulated tend to
have higher productivity.
Aghion et al. (2003b) and Aghion et al. (2004) are of particular relevance to
our results. They assess the impact of the institution of minimum wage laws on
firm-level investment and use the 1991 trade liberalization in India to illustrate how
reform may have unequal effects on industries and regions. Their main theoretical
results are that liberalization enhances investment in industries that were initially
close to the technological frontier and that pro-worker legislation lowers investment
and this negative effect is magnified by liberalization. Their empirical results confirm
the main predictions of the model.
This dissertation contributes to this strand of literature in two ways. Firstly,
50
most theoretical models do not try to model industrial policy in India and how
this affected market structure and hence innovation incentives. The policy of close
control of private sector initiative was in place for nearly 40 years and affected
all aspects of production, distribution and even consumption. Our model tries
instead, to capture the effect of policy on market structure and the predictions of
our theoretical model provide a richer characterization of innovation incentives.
In addition to this, we address the question of whether our results regard-
ing domestic de-regulation are eroded or enhanced as a result of competition from
high technology foreign firms. That is, what is the interaction between domestic
competition and foreign competition?
Secondly, contrary to both Aghion et al. (2003b) and Aghion et al. (2004)
who assess the impact of deregulation on industry productivity we use detailed and
reliable firm-level data for the 15 year period. Most other studies on India use
industry-level data. Further we have a detailed (and to our knowledge, the only)
data set on Indian industrial policy from 1970 to 1994. Instead of analyzing the
impact of isolated reforms (for example Aghion et al. (2004) use deregulation in
1985) we are able to analyze the impact of all changes in industrial policy in the
1980s and 1990s. We provide more details about our data in Section 3.4.
As in the case of increasing competition, theoretically trade liberalization could
have both a positive as well as a negative impact on productivity. Different theo-
retical models give startlingly different results about impact of trade reforms. Thus
the effect of trade liberalization on productivity is an empirical question. While
the early evidence was somewhat mixed (Tybout (1992)), recent surveys by Tybout
51
(2000) and Epifani (2003) conclude that the empirical literature generally support a
positive effect of trade liberalization on productivity. For the case of India, several
studies have used micro data to study the impact of trade reform on productivity
((e.g., Krishna and Mitra (1996), P. Balakrishan and Babu (2000)) and have come
to very different conclusions. Krishna and Mitra (1996) who use firm-level data for
the period 1986-1993 from several industries and find strong evidence of an increase
in competition (as reflected in the reductions in price-marginal cost markups) and
some evidence of an increase in the growth rate of productivity. P. Balakrishan and
Babu (2000) use a sample of 2300 firms over the period 1988-89 to 1997-98 to find
no evidence of acceleration in productivity growth as a result of the 1991 reforms.
Das (2003) uses standard growth accounting on manufacturing industries during
1980-2000 and finds no evidence of change in TFP growth following the 1990s re-
forms. More recently, Topalova (2003) has used a more sophisticated methodology
for calculating productivity and concludes that reductions in trade protection lead to
higher levels of productivity in Indian manufacturing over the period 1989-2001. In
a recent study Sivadasan (2003) uses a comprehensive and highly detailed firm-level
data set and finds an 30 to 35% increase in mean intra-plant productivity level in
tariff liberalized industries. There is also a 25% increase in aggregate output growth
and a 20% increase in aggregate productivity growth following tariff liberalization.
We contribute to this literature by providing reliable estimates of the impact
of trade reforms on firm-level productivity after controlling for the effect of domestic
de-regulation. Other than Aghion et al. (2004) who use limited deregulation data for
one year of the 1980s, none of the papers mentioned above control for the industrial
52
de-regulation that took place in India during the 1980s. Hence their estimates of
the impact of the trade liberalization of 1991 may be biased since some of the rise in
productivity may have been due to de-regulation during the 1980s. Interest in the
reforms of the 1980s has recently been revived due to a puzzling empirical result.
As Panagariya (2002), Panagariya (2004) and Delong (2001) point out, reliable
productivity measures show a sharp rise in productivity levels and rates of growth
prior to the 1991 sweeping reforms. In a recent paper Rodrik and Subramanian
(2004) argue that the structural break in Indian growth came in the early eighties
because there was an attitudinal shift on the part of the government towards a
pro-business approach rather than due to actual policy changes like de-licensing.
However, unlike us they do not use data on policy reform to prove this hypothesis.
3.3 Background on Indian Industrial Policy
The licensing of industries was one of the major methods to control private enterprize
in India. The mixed economy framework mandated a role for the private sector .
However, it was felt that the private sector would need encouragement to invest
in the desirable areas. Hence industrial licensing evolved as a method to direct
investment in desirable directions. Details about the evolution of the system are
presented in Appendix B.1. Here we present a brief description of the system and
important changes.
A license was a document that permitted a firm to continue/begin production
in an industry. It was issued by the Ministry of Industry in New Delhi. Under
53
the Industries (Development and Regulation) Act 1951 (henceforth referred to as
IDRA), all factories (defined as enterprises that did not use power but employed
more than 100 workers or enterprises that used power and employed more than 50
workers) that were already operating or wished to operate in a specified list of
industries were required by the government to obtain a license.
The scope of a license was fairly broad especially from the late 1960s onwards.
Almost by definition, the licensing regime controlled entry into the industry and
hence the amount of competition faced by a firm. Section 1 of Appendix B.1 goes
into details about the considerations of the licensing committee while debating a
license. These were mainly macro-economic in nature and had little to do with
the merits of the project. A license also specified the amount of output that a
firm could produce. It was conditional on the proposed location of the project.
Permission would be required to change location. The exact nature of the item to
be produced was also specified and the firm needed to take permission or another
license to change his product mix. Even the kind of technology and inputs that the
firm could use in production (though not specified on the license) was determined
because the most crucial raw materials (steel, cement, fuel etc) were controlled by
the government and the firm needed to get annual allotments of these for production.
However, the effect of a permanent license to produce (combined with a low
threat of potential entry) on firm-level incentives to reduce costs, modernize tech-
nology, improve quality and engage in monopolistic practices was raised by one
committee after another starting from 1965, a mere 15 years after licensing was
implemented. One of the earliest observations were made by the The Monopolies
54
Inquiry Commission 1965 chaired by K. C. Dasgupta.
?.....the requirement of law that new industries with capital over a specified
amount could not be started without a license is a formidable obstacle in the way of
new entrepreneurs freely entering the lists.? 1.
Further,
?The system of controls on the shape of Industrial licensing however necessary
from other points of views, has restricted the freedom of entry into industry and so
helped to produce concentration?2.
Thus, the licensing regime in India affected firm-level productivity and costs
through its control on both the firm?s ability and incentives to innovate, reduce costs,
adopt new technology etc. The direct controls on outputs and inputs affected ability
and the indirect control of entry affected incentives. Even if the direct controls were
not implemented fully due to corruption etc, the effect of the indirect controls on
incentives was very large. Licensing restricted entry into most sectors and created
artificial monopolies and oligopolies. The average four-firm concentration ratio in
Indian manufacturing in 1981 was 54.2% compared to 32% for the US in 1977. Even
among developing countries, India seems to be closer to Poland (64.8% in 1988) than
Brazil (32% in 1988).
1Dasgupta (1965), page 7
2Dasgupta (1965), page 8
55
3.4 Data
The data used in this paper comes primarily from two sources. In order to measure
changes in the competitive environment faced by a firm (the right hand side explana-
tory variable), we have collected a detailed data set of industrial policy in India (to
our knowledge, the only one existing) from the 1970s onwards. Using this data,
we can identify which four-digit industry underwent reform (freedom from licensing
requirements) in each year of the 1970s, 1980s and 1990s. The main source of data
was internal government publications and notifications . Some commonly available
publications like the Economic Survey were also used. Common publications, how-
ever, do not reveal the level of detail about the conditions under which firms were
eligible to avail of certain policies. Study of government notifications, memos etc
provided insight into the ideology of policy-makers and provided rich detail that we
exploit in our identification strategy.
It is important to note that this is the first attempt to collect and compile
data on Indian industrial policy for the period 1970-1995 in a manner that allows
econometric estimation of the impact of policy changes. One reason for this is that
detailed data on industrial policy pre-1991 are available only in obscure government
notifications which are not easily available. Secondly, in the literature on India, the
reforms of the 1980s are usually dismissed as too small and scattered to have had a
significant effect on anything. There has not been a single detailed empirical study
of these early reforms. As mentioned earlier, the reforms of the 1980s have been of
great interest lately. Reliable productivity measures for India show a sharp rise in
56
productivity levels and rates of growth prior to the 1991 sweeping reforms.
In the 1980s, the government started relaxing the licensing regime by ?de-
licensing? certain industries. From the late 1960s onwards it was starting to get
clear that the strangle hold of regulation on Indian industry was fatal for it and
many assessments of the system in the 1960s and 1970s advocated relaxation of
regulations. Certain attempts were made in the 1970s 3. But it was in the 1980s
that any significant change in the working of the system occurred. Table 3.1 shows
the percentage of manufacturing output and value added that was de-licensed in
each year of the 1980s and 1990s. We also show the percentage of factory output
and value added that was de-licensed ( factories are defined as firms not using power
and 100 or more employees and firms using power employing 50 or more employees)
since under the IDRA 1951, only these factories were under the ambit of licensing.
This piece-meal approach to reforming industrial policy continued through
the 1980s. In 1991, the Indian economy faced a balance of payment crisis and
was forced to take loans from international organizations. Under pressure from
these organizations, the biggest de-licensing episode occurred. Almost all industrial
licensing was removed (other than for 16% of manufacturing output). Along with
this, there was across the board reductions in tariffs and rationalization of non-tariff
trade barriers. The rupee was de-valued by 22% (from Rs. 21.2 against the dollar
3The first ?de-licensing? of firms was done in 1966 and by 1969, 41 items were de-licensed.
However in 1973, these industries were licensed again and it was only in 1975 that the second
phase of de-licensing began when 21 industries accounting for 3% of manufacturing output were
de-licensed.
57
to Rs. 25.8). The sheer scale and scope of the reforms were so large that this reform
episode has been the one that has caught the imagination of policy-makers and
researchers alike.
Table 3.1 brings forward two important points that challenge the predictions
of other studies on India. The first is that with respect to the percentage of manu-
facturing output that they affected, the reforms of the 1980s were quite significant.
Cumulatively, 23% of output and 23.6% of value added in 1990 had been de-licensed.
Hence, studies that ignore pre-1991 changes in the licensing regime provide mislead-
ing estimates of the impact of the 1991 crisis. Secondly, de-licensing in 1991 was
not ?across the board? as is the common assumption in most studies. 16% of man-
ufacturing output and value added were still under compulsory licensing post-1991.
Some of these industries were gradually de-licensed in 1993 and 1994. But studies
that ignore the actual chronology of de-licensing post-1991 also overstate the impact
of the 1991 crisis.
To measure the left-hand side variable-productivity of the firm- we use a rich
and rarely used unit-level database. The source for unit-level data for this study is
the Annual Survey of Industries conducted by the Central Statistical Organization
(CSO), a department of the Ministry of Programme Planning and Implementation,
Government of India. The survey covers all factories registered under the Factories
Act 1948 (defined as units employing 20 or more workers). Note that the survey cov-
ers only the formal sector in Indian manufacturing. The ASI frame can be classified
into 2 sectors-the census sector and the sample sector. Units in the census sector
(all factories with more than 100 workers) are covered with a sampling probability
58
of one while units in the sample sector are covered with probabilities one-half or
one-third.
This rich unit-level database has been rarely used in literature since it has only
recently been made available to researchers in electronic format. A recent paper that
has utilized this database is Sivadasan (2003) who calculates the impact of the trade
and foreign investment reforms that were done in 1991 on total factor productivity
in Indian manufacturing. Most other papers (for example, Aghion et al. (2004)) use
ASI data at the industry-state level.
We obtained this data for the 14 year period 1980-81 to 1994-95. The length
of our data allows us to cover all the reforms of the 1980s as well as the major reform
episode of 1991. The data is reported on a financial year basis. That is, 1980-81
refers to the period between 1 April 1980 and 31 March 1981.
For our analysis, we use all unit-level data on all factories as defined by the
Industries (Development and Regulation) Act 1951 since these units were the ones
that came under the ambit of industrial licensing. These are units that employ 50
or more workers and use power in the production process or units that employ 100
or more workers without the aid of power. The non-factory units that we exclude
are large in number (an average of 30% over our time period) but account for only
14% of total output and 18% of employment in registered manufacturing.
Since we use log of output per worker, observations for which the output or
worker figures are less than or equal to zero are excluded from the analysis. Further,
42000 firms are listed as being closed in the year of the survey and these are also
excluded from the analysis.
59
The basic characteristics of the data that we use for our analysis are summa-
rized in Table 3.2. As mentioned earlier the census sector of the ASI comprises of
units employing 100 or more workers while our subset consists of units employing 100
or more workers without power and 50 or more workers with power. Thus, nearly
80% of our data has a sampling probability of one. On average there are around
15000 units a year corresponding to a population of 20000 units. Note that the
sampling scheme changed in 1987. In all our analysis, we weight observations using
the multiplier or the inverse sampling probability to adjust for sampling frequency.
Summary statistics of our key variables are presented in Table 3.3.
Real output is measured as value of total output measured in 1993-94 prices.
The deflator used is the sector specific wholesale price index 4. Labor is measured
as the number of employees.
3.5 Identification Strategy and Estimation Equation
3.5.1 Estimation Equation
The prediction from our theoretical model that we want to test is that a more com-
petitive domestic environment forces firms to raise their investment in productivity-
enhancing technology and hence leads to higher productivity. As explained in Sec-
tion 3.3 the industrial licensing regime in India affected market structure and market
power of firms operating in the manufacturing sector, mainly through restriction of
entry. It also affected investment incentives of firms via its control of firm output,
4Source: various issues of the Reserve Bank of India Bulletin
60
location, raw material usage etc. Thus, a removal of licensing requirements from an
industry led to a more competitive, market-driven environment. The main equation
that this translates into is given by
yijts = ? +?1Dejt +?Xijts +?ijts (3.1)
Here yijts is the log of output per worker for a factory i producing in industry
j, located in state s at time t. Dejt is an indicator that takes in a value of one in
all years greater than equal to year t if industry j was de-licensed in year t.
However we need to control for macroeconomic changes that may affect firms
for example rates of interest, exchange rate etc. We control for these by putting
in year fixed effects into our regression equation. The second area of concern is
industry-specific factors that may affect firm-level productivity and we control for
these using industry-fixed effects. Thus our estimation equation is
yijts = ?0 +?j +?t +?1Dejt +?Xijts +?ijts (3.2)
Here ?j, ?t are industry and year effects. The vector Xijts contains other firm,
industry and state level controls. We explain these controls in detail in Section 3.5.5.
3.5.2 Identification Problem and Strategy
The estimation of Equation 3.2 may not give unbiased, consistent estimates. Note
that main explanatory variable Dejt varies at the industry level. Political economy
factors like political affiliation, lobbying power etc are also at the industry-level and
these might affect whether industry j gets de-licensed. Since these political economy
61
factors are not observed, they become part of the error term. But this then means
that the error term is correlated with a key right hand side variable- the de-licensing
dummy- leading to omitted variable bias. Under these circumstances we will not be
able to identify the co-efficient on Dejt. Industry-fixed effects may capture some of
these political economy un-observables. However, in order to get reliable estimates
of the impact of industrial de-regulation we need within-industry variation.
One important reason for our concern is that the reforms of the 1980s have
been characterized by some as ?reforms by stealth?. There was no consensus for
economic reforms in the 1980s. It is clear from policy documents that the govern-
ment was at pains to portray the changes of the 1980s as a continuation of the
existing system (even though these were dramatic changes that veered away from
the high-regulation, socialist paradigm in operation). Under these circumstances it
is possible that the government was choosing industries for de-regulation based on
certain characteristics that either raised the chances of the success of the reforms (for
example, picking high productivity industries) or that minimized social costs in case
of failure (for example, picking high technology industries to minimize employment
effects).
Our identification strategy uses important institutional features of industrial
policy in India. Firms that had assets in plant and machinery, land and building
less than a certain amount were free (?exempt? in official parlance) from indus-
trial licensing requirements. This was an important component of the governments
strategy to promote the growth of smaller firms that would be main engine of em-
ployment generation in the manufacturing sector. These firms would also satisfy
62
consumption needs of the economy and would be a mechanism to spread economic
growth to far-flung regions of the country. Further, this was also a pragmatic deci-
sion so that the administrative burden of the licensing regime did not grow too large.
By the late-1960s, as the manufacturing sector took off, there were reports of long
delays in the grant of licenses. By giving exemption to some firms, the intent was to
lessen the sheer volume of applications that the licensing authority had to process.
Note that these exempt firms could enter and operate in any industry and without
restrictions on output or investment as long as they stayed below the threshold 5.
This meant that in each licensed industry there is a group of firms that is not under
licensing. We call these the ?exempt? firms.
Further, in the 1980s certain industries were ?de-licensed?. That is, all firms
(irrespective of whether they were exempt or not) were freed from licensing re-
quirements in these industries. In terms of our identification strategy, this means
that from 1984 onwards all firms in certain ?de-licensed? industries were not under
licensing requirements.
Keeping in mind the institutional features of industrial policy in India, in each
de-licensed industry there is a group of firms that are not affected by the reform since
they were not under licensing to begin with (the exempt firms). So the industry
in which a firm produces as well the firm?s exemption status jointly determine the
5In some cases, certain additional conditions like upper limits on foreign exchange requirements
for the project needed to be fulfilled in order for the firm to be ?exempt?. Despite this, according
to officials, a large number of firms were able to take advantage of this scheme. However note that
this means that entry into exempt firms was not entirely free.
63
firm?s exposure to de-licensing.
To sum up, a firm is treated by de-licensing reform in industry A in year t if
? Industry A was de-licensed in year t.
? The firm was under licensing to begin with. That is, it was not granted an
exemption from licensing.
The criterion for exemption was in terms of the original rupee value of plant
and machinery, land and building owned (or proposed to be owned) by the firm.
This definition was constant for all industries. That is, an exempt firm in industry
A would face the same threshold as an exempt firm in industry B. Over our sample
period 1980-1994, the rupee value of this definition was changed twice-in 1983-84
and 1990-916. From the period 1980-81 to 1982-83, a firm whose assets in plant
and machinery were worth less than of equal to Rs. 30 million was defined as small.
In 1983-847, the threshold was increased to Rs. 50 million and in 19908, to Rs. 150
million. Note that inflation in the price of land or machinery would not change the
exempt/not-exempt status of the firm since this was based on the original or book
value of assets (not current value).
However, this ?exemption? from licensing was not available to firms in a group
6In May 1990, in an Industrial Policy statement the government raised the exemption limit to
Rs. 500 million. However, in November 1990 a new government came to power and by April-May
1991, the foreign exchange crisis was taking hold. Given these two factors, the implementation of
the May 1990 statement in actual practice is in doubt.
7Source: Ministry of Industry Notification S.O. 328(E) dated 23 April 1983.
8Source: Economic Survey 1990-91.
64
of industries referred to as Schedule IV and Schedule V industries 9. Schedule
IV included industries like textiles produced on power looms that were items of
mass consumption over whose production, pricing and distribution the government
wanted close control. Schedule V industries were the industries that were deemed to
be using ?scarce? raw materials (imported or domestic). So in these industries, all
firms were under licensing. We do not include these industries in our main estimation
equation but we use them to provide a test of our identification assumption.
Thus the actual equation that we estimate is
yijts = ?0 +?j +?t +?1Dejt +?2NotExemptit +?3DejtNotExemptit +?Xijts +?ijts
(3.3)
Here NotExemptit takes on a value of one when the firm is not exempt
from licensing requirements. Our identification strategy allows us to deal with
the problem that industry-level un-observables may make Dejt correlated with the
error term. If we suppose that there were unobservable factors that determined
whether industry A was de-licensed in year t, as long as these factors were industry-
specific they should be same for all firms within industry A. That is, a exempt firm
(NotExemptit=0) in industry A has the same un-observables as does a non-exempt
firm (NotExemptit=1) in industry A. Thus, any variation in productivity around
9Source: Government of India, Ministry of Industry Notification 16 Feb, 1973. S.O.
98(E)/IDRA/29B/73/1 specifies these industries and the subsequent changes to these lists were
gathered from amendments to this notification as well as other notifications and press notes issued
by the government.
65
the threshold for licensing should come from de-licensing. Thus, co-efficient ?3 on
the interaction of size and de-licensing -DejtNotExemptit- is identified.
As mentioned in Section 3.3, the licensing regime affected both the ability as
well as the incentives of a firm to invest in productivity. The controls on the amount
of output, the location of the plant, the technology used etc directly controlled the
ability of a firm while controls on entry into an industry affected incentives. When
an industry was de-licensed, both these controls were removed simultaneously. In
this context, the coefficient ?3 captures the impact of the relaxation of both direct
and indirect controls on firms. That is, ?3 measures the impact of de-licensing on a
non-exempt firm?s productivity because of relaxation of output, locational and other
constraints as well as the impact on productivity due to more competition and a
higher threat of potential entry.
The first mechanism?relaxation of direct controls on the ability of a firm?
makes sense since only the not exempt firms were under the burden of fulfilling
onerous conditions on output, location etc while exempt firms were not (as long as
they maintained assets below the government definition). Thus, there is differential
impact on these two types of firms within an industry of de-licensing.
The argument with regard to the second mechanism?higher threat of potential
entry?is more nuanced. One might argue that if there were no barriers to entry as
a small, exempt firm while there were huge barriers to entry as a large, not exempt
firm (in the form of the cost of procuring a license) then no firm would want to
enter as a not exempt, large firm in the pre-reform period. All firms would enter
small and would grow until they reach the threshold amount of assets. This means
66
that de-licensing will not increase entry into the ranks of not exempt firms while
exempt firms have free entry both pre- and post-reforms. If this were the case then
our identification strategy does not capture the impact of entry de-regulation on the
productivity of firms.
However, the argument above assumes that there are no costs to entering as
a small, exempt firm. This assumption is not likely to be true because a license
to produce was only one in a whole package of permissions and permits that a
firm needed to get in order to commence production. For example, in order to get
exemption firms needed to agree to an upper limit on foreign exchange requirements.
Further, on the basis of the recommendations made by the licensing committee, the
firm had to get allotments of essential raw materials from another committee that
was in charge of allocations. Similarly, the firm needed to go through a separate
procedure to get permission to procure foreign exchange and to import any raw
materials or machines. Further, the firm needed to get financing for its project but
applied to financial institutions after getting all the required permits and licenses.
All these procedures and permissions were more likely to be more difficult and costly
for a small firm. So if a firm decided to enter the industry as small, the firm not only
suffered the loss of economies of scale but also faced additional costs of being small.
Thus, while there were no explicit entry barriers into the exempt category of firms,
the other accoutrements of the licensing and trade regime created entry barriers
and it is by no means obvious that a firm would always want to enter small in size,
pre-reforms. This means that de-licensing of an industry reduced entry barriers for
both of exempt as well as not exempt firms and that our empirical strategy catches
67
these effects.
Our identification strategy has been used extensively in the evaluation of social
programs in developing countries. Similar strategies have been used in the public
finance literature to evaluate the effect of public policies. Rosenzweig and Wolpin
(1988) first proposed the use of fixed effects methods for evaluating the impact of
public programs in developing countries. Gertler and Molyneaux (1994) applied
these methods to the problem of estimating the returns to family planning pro-
grams by controlling for unobservable region-specific cultural and religious norms.
Duflo (2000) uses this strategy to evaluate returns to a school-building program
in Indonesia. She uses variations in program intensity over regions as well as the
date of birth to identify individuals who would be most affected by this program.
Children who were at or below the school-going age at the time the program was
implemented will benefit more than children who were older. Further, this differ-
ence is higher in regions that received more schools. Angrist and Evans (1996) use
a similar strategy to identify the effects of state abortion reforms on labor mar-
ket outcomes. The causal relationship they want to identify is between teen-child
bearing and labor market outcomes. However, it is possible that women who bear
children as teenagers would have bad labor market outcomes due to unobservable
variables leading to biased OLS estimates. So Angrist and Evans (1996) create an
instrument for teen child-bearing that is an interaction of the state of birth and
the year of birth. The latter captures the exposure that a woman had (in her teen
years) to liberal abortion environment. So women with more years of exposure to a
liberal environment are less likely to be teen-mothers than women with less years of
68
exposure. However this difference is likely to be more in states that reformed their
abortion laws. More recently, Visaria (2004) uses this strategy to identify the effect
of judicial quality on economic outcomes. The sources of variation in her applica-
tion are the monetary threshold for claims to be eligible for these tribunals, and the
staggered introduction of debt-recovery tribunals across Indian states.
The Indian reforms of the 1980s are well-suited to the use of fixed-effects meth-
ods because the variations in the regulatory environment come from a well-defined
series of reforms. This makes it possible to test an implication of the identification
assumption that any variation in performance of firms due to de-licensing should
come around the actual exemption status. We use a group of industries where all
firms were under licensing i.e. no firm was granted an exemption. For this group of
industries, since the threshold for exemption did not apply, the co-efficient on the
interaction of Dejt and NotExemptit should be insignificant. That is, there should
be no variation in the response to de-licensing around the exemption threshold. The
results of this specification test are reported in Appendix B.3.
3.5.3 Exogeneity of Exemption Status
In order to get unbiased and consistent estimates from Equation 3.3, we need to make
sure that the sources of variation in our data are exogenous. The first issue was that
of omitted variable bias that may make the de-licensing dummy endogenous. To
solve that we use an institutional feature of Indian policy making, as explained in
the previous section.
69
However, a second source of bias arises from the possibility that firms may be
endogenously choosing their exemption status (i.e. the size of their assets in plant,
machinery, land and building) based on anticipations of de-licensing. Suppose that
the main benefit to a firm for choosing a low level of capital comes from the fact that
the firm is free from licensing requirements. Then if the firm is truly constrained by
the threshold and is anticipating de-licensing of its industry (when it will loose its
privileged status) it will increase its capital even if that means that it is above the
threshold.
There are several reasons why an exempt firm will want to increase its capital
if it anticipates de-licensing. The first is that if the main reason why the firm is
keeping a low capital stock is to escape from licensing then as soon as this privilege
is removed, the firm will no longer benefit from constraining its capital and hence
will want to expand. Another reason is that de-licensing means that the firm will be
competing in an environment of free entry for larger firms (remember that exemption
was only for ?smaller? proposals). In order to compete it might be necessary to
take advantage of economies of scale and that might entail increasing its assets.
In the sections below we investigate the possibility that firms chose their assets in
anticipation of de-licensing of their industry. If this is the case, there should be
a difference in the distribution of firms over industries. Industries that were de-
licensed should have a lower proportion of exempt firms (as these firms raise their
assets in plant, machinery, land and building because they no longer stand to gain
from having low assets). Further, the rates of growth and levels of assets of exempt
firms should be higher in de-licensed industries post-reform.
70
The first important point to note about de-regulation reforms in India is that
there was an exogenous break in the pattern of de-regulation in 1991. Up to 1991,
de-licensing was done in a piece-meal manner-a few industries each year. The extent
of the balance of payment crisis in 1991 and the conditions attached with foreign
loans were completely exogenous. By far the largest de-licensing episode occurred in
1991. Thus, it seems unlikely that firms would be choosing their assets endogenously
with respect to the de-licensing in 1991. Thus, in our discussion of endogeneity of
exemption status we focus on de-licensing episodes in the 1980s.
With regard to the reforms in the 1980s, we conduct a number of tests of
the exogeneity of exemption status, in terms of size of assets. An analysis of the
data reveals little evidence of firms choosing their assets (and hence their exemption
status) in anticipation of de-licensing.
(a) In the industries that were de-licensed during the 1980s, the proportion of ex-
empt firms was 0.95 pre-reform (1980-83) and it fell to an average of 0.93
during the reforms (1984-89). This pattern was identical to that in industries
that were not de-licensed in the 1980s (here the proportion of exempt firms
fell from 0.97 to 0.95). That is, there was no sudden inexplicable decline in
the proportion of exempt firms post-reform. Similarly, the average annual rate
of growth of the number of exempt firms also maintained its pattern across
industries (It rose from 1.6% to 2.4% in the de-licensed industries and from
-1.5% to 0.5% in the still-licensed industries.).
(b) In Table 3.4 we investigate the pattern of the average annual rates of growth
71
of assets in plant, machinery, land and building across exempt firms. We see
that while the rate of growth of the mean is higher in de-licensed industries as
compared to licensed industries, the top 25% of firms grew faster in licensed
industries than in de-licensed industries during 1984-89 (12.97% as compared
to 10.53%). This is contrary to what we would find if firms were choosing
their assets based on reforms. Similarly in the top 50% of firms, the rate
of growth of assets is slightly higher for exempt firms in licensed industries.
Thus, in firms that were closer to the threshold-and hence more likely to be
endogenously choosing size- there is little or no difference in the behavior of
growth of assets.
(c) In Figure 3.1 we investigate the distribution of exempt firms over assets in
industries in two year-1983 (the year preceding the first deregulation episode)
and 1988 (the year immediately after the reforms of the 1980s). Within each
graph the solid line represents the year 1988 while the dashed line represents
the year 1983. We see that the distributions are very similar in de-licensed
industries as compared to licensed industries. This is especially true towards
the end of the distribution i.e. nearer to the threshold.
Note that there is a large mass of firms with very little assets in plant, ma-
chinery, land and building. To get a clearer idea of the effect of reforms on the
distribution of assets, in Figure 3.3 we present the distribution assets of firms
with assets below Rs. 4.5 million (ie. the really small firms) and in Figure 3.2
the distribution of assets of firms with assets greater than Rs. 4.5 million but
72
less than the threshold amount of Rs. 50 million10 . We see that the distribu-
tion of really small firms (Figure 3.3) in de-licensed industries is quite different
from the distribution in licensed industries especially in the lower range. But
this difference is constant over time i.e. it holds for both years. Towards the
end of the distribution, in both types of industries, the density is higher in
1988 as compared to 1983.
In Figure 3.2 again we see that the distribution of assets is very similar in
licensed and de-licensed industries especially as assets rise. Towards the be-
ginning of the distribution (i.e., for firms with assets close to Rs. 4.5 million),
licensed industries tend to have a higher density in 1988 while de-licensed in-
dustries tend to have lower density. But the important point to note is that
this difference crops up very far from the threshold of Rs. 50 million. For firms
closer to the threshold and hence more important for us with regard to the
potential problem of endogeneity, the distributions are very similar.
Thus we see that the behavior of exempt firms with respect to assets in plant
and machinery, land and buildings is not different over de-licensed and licensed
industries11. Hence the argument that firms may choose size in anticipation of de-
10We choose the threshold of Rs. 4.5 million since this categorization was used by the government
to define ?small? firms. These firms were given special assistance and encouragement from the
government. In this respect it is interesting to analyze the behavior of these firms. We also
analyzed the behavior of exempt firms around randomly chosen thresholds and our findings are
similar. There is little difference in the behavior of exempt firms (no matter how far from the
actual threshold) across licensed and de-licensed industries.
11We also analyzed the distribution of other firm characteristics like assets in plant and machinery
73
licensing does not seem to hold in our data.
(d) Other evidence against this argument arises from an analysis of whether the
threshold (on assets in plant, machinery, land and building) implemented by
the government was binding for exempt firms. If there was a large mass of
exempt firms that were on or slightly below the threshold and were maintaining
low levels of assets deliberately then the magnitude of the endogeneity problem
is bound to be quite large. However as we demonstrate in Appendix B.2,
we do not find any evidence of bunching below the threshold for licensing.
We do not find a large mass of firms directly below the threshold12. The
threshold is binding only for the top 1% of firms (that too with a utilization
rate of only 80%) and utilization rates drop of sharply for the other percentiles.
Further, there is no evidence to suggest that firms directly above the threshold
(and hence, licensed) were racing to raise their assets in plant and machinery-
another indication that the asset limits might not be constraining13.
and employment over exemption status and industries. Here too we find that the patterns are very
similar in de-licensed industries as compared to licensed industries.
12In Appendix B.2 we also conduct this analysis for the capital to labor ratio and find that there
is no bunching of firms around the average capital-labor ratio (as implied by the threshold) within
an industry and year.
13This point should not be taken to imply that the licensing regime itself was not binding since
firms were not trying to escape it. As pointed out in Section 3.3, the licensing system controlled
two things at the firm-level-a firms ability (via control of the amounts of capital, output etc) and
a firms incentives (via control of entry into the industry).
74
3.5.4 Comparability of non-equivalent groups
Our empirical strategy depends on the difference in performance between exempt
and not exempt firms in response to de-licensing. However exempt and not exempt
firms are different in a crucial dimension? size. Hence we are concerned that there
may be other variables, other than de-licensing that may lead to this differential
response between exempt and not exempt firms. In order to check the equivalence
of these two types of firms, we conduct a series of simple tests in Appendix B.5. We
find that in terms of trends of major variables as well as the evolution of the distri-
bution of major variables, not exempt and exempt firms behave in a similar manner
particularly in the pre-reform period. That is, there are differences in exempt and
not exempt firms but that these differences do not change over time or due to any
other variables prior to the reforms.
3.5.5 Controls
In this section we talk about the controls that are included in the Xijt variable in
Equation 3.3. We use our industrial policy database to control for policies that may
affect the ability and/or incentives of the firm to raise its productivity.
? Policies regarding dominant firms and firms belonging to large and/or
foreign owned industrial houses. Control of large private industry was an
important objective of policy in India. The main objective was to prevent the
concentration of economic power in a few private hands. Since the inception of
the licensing system, certain large industrial houses were considered dangerous
75
and hence all licensing proposals from these houses were treated with suspicion.
In 1973, the government streamlined its policies regarding large industrial
houses and announced a list of industries in which large and/or foreign owned
houses were allowed to operate14. We control for this list of industries in which
large and/or foreign owned houses were allowed to operate. We define
Largejt =
?
????
???
?
1 if industry j was open for large houses in year t
0 otherwise
? Controls for the ownership and organization structure of the firm. Con-
trols for these are included since the ability to take productivity enhancing
decisions and the flexibility that a firm has in adapting to a changed policy
environment may depend on whether the firm is owned by the government
or by the private sector. Similarly, joint family owned firms might behave
differently compared to private limited firms or co-operative societies. Fur-
ther, ownership and management structure of the firm might affect financial
constraints faced by the firm and hence the assets of the existing or proposed
factory. We define dummy variables for each type of ownership and organiza-
tion structure15.
14Even in these industries, the stated policy was one of preference for non-large firms. If a large
house wanted to set up a firm in an industry not on this list then it would have to undertake an
export obligation of 60-75% of output.
15For ownership, there are 6 categories-owned by central government, owned by state govern-
ment, jointly owned by state and central government, wholly privately owned, joint sector firm
with majority private ownership and joint sector firm with majority public ownership. Organi-
zation includes individual proprietorship, joint family run, other partnership, public limited firm,
76
? The location of a firm is another important variable that maybe correlated
with its productivity or its size. In addition to controlling for the state in
which a firm is located, we also control for whether it was located in an urban,
rural or metropolitan area. It is possible that the level of infrastructure that
are available to a firm is different if it is located in an urban area as compared
with a rural area and this in turn maybe correlated with the productivity of
the firm. We define urbani = 1 if firm i is located in urban area, 0 else and
metroi = 1 if firm i is located in metropolitan area, 0 else.
? One issue that we are concerned with is the relationship between the size
of the firm and the productivity of the firm. In particular, productivity
of the firm as measured by output per worker may be determined by its size.
If we were to measure size by output then this would be true by definition.
Our definition of size however, is based on the assets of the firm in plant
and machinery. But it is still possible that output per worker (our measure
of productivity) is affected by capital per worker (for example, in any linear
homogenous production function) and so size (and hence exemption status)
and productivity are directly related. We cannot directly put the capital-
labor ratio in our estimation equation (since the variable Neit is a function of
capital). So we need a proxy for the capital to labor ratio in our estimation
equation.
We use the average price of capital relative to price of labor in a state s in year
private limited firm, public corporation, co-operative society and other.
77
t. We use the deflator for net fixed capital formation (with base 1993-94)16 as
a proxy for the price of capital. In our unit-level data we have the wage bill
as well as the number of workers employed by each unit. So we are able to
impute the annual wage rate that a firm has to pay. Since this wage rate is in
nominal terms we deflate it so that it has base 1993-94 (similar to our measure
for price of capital). The reason why we use state-year averages of this cost
ratio is because firm level cost ratios might be endogenous with technology and
productivity choices of the firm. Further, labor laws (for example, minimum
wage) and their implementation is at the state level. So the average cost ratio
captures the major institutional features of the labor market in India.
3.6 Results
3.6.1 Main Results
In Table 3.5 we present our baseline results for the estimation of Equation 3.3.
In each specification we have included all the controls mentioned in the previous
section. These regressions are within-industry. That is we are estimating a firms
performance relative to the average performance of the 4-digit industry to which the
firm belongs. We have also included year and state fixed effects. Standard errors are
clustered around four-digit industry. Column 1 of Table 3.5 shows the results of our
main specification on the sample of factories, excluding Schedule IV and Schedule V
industries. Thus we see that on average firms that were not exempt from licensing
16Source: National Account Statistics
78
and were in de-licensed industries had 16.1% higher labor productivity. Our estimate
is significant at the 5% level. That is, industrial de-regulation leads to a significant
rise in the firm productivity. As mentioned earlier, licensing affected both the ability
and the incentives of a firm to become productive. Hence, the 16.1% increase in
productivity is due to removal of direct restrictions on firms (an impact on the
ability of the firm) as well as due to higher competition and lower entry barriers
into the industry (an impact on the incentives of a firm).
In Column 2 of the table we present regression results for the period 1980-90.
Note that an important point of our analysis is that the reforms of the 1980s have
been unfairly ignored by the literature. So it would be interesting to see the impact
of those piece-meal reforms on productivity, not the path-breaking reforms of 1991.
We find that the average firm that was treated with de-licensing during the 1980s
did better than a non-treated firm. However this effect is significant only at the 17%
level. Thus, the reforms of the 1980s did have some impact on firm productivity
but this effect was small. Note that the total impact of de-licensing (?1+?3) is very
small for the period 1980-90. One possible rationalization for this small impact on
productivity during the 1980s could be the lack of an enabling environment. That
is, a rise in domestic competition can change the incentives of the firm to raise its
productivity but this effect is reinforced by other factors. One such important factor
that we will investigate is competition from abroad i.e. trade reform.
However, different industries may have very different productivity paths and
that our coefficient may be capturing some of this time-varying heterogeneity. In
Column 3 we check the robustness of our results to the inclusion of 3-digit industry
79
specific time trends in addition to year effects. In Column 4 we report the results for
the year 1980-90 after including industry-specific trends. We see that the interaction
term is considerably lower in magnitude and is significant at the 10% level. Thus,
even after controlling for time-varying growth paths in different industries, we find
that firms that were affected by de-licensing had higher labor productivity than
firms that were not affected. For the 1980s, the coefficient on the interaction is
positive and insignificant.
Thus our baseline results support our intuition and the first prediction from
our theory-firms in more competitive, deregulated industries tend to have higher
productivity. An important point to note from Table 3.5 is that in all the specifica-
tions, the average impact of de-licensing on all firms in the industry (the coefficient
on Dejt) is negative. That is, industries that were deregulated have lower produc-
tivity on average compared with industries that were not deregulated. It is the not
exempt firms (that were under licensing) that see a rise in their productivity after
de-licensing. But in almost all cases, the total effect (?1 +?3) is positive17.
From Table 3.5, the role of exemption status also becomes clearer. In all the
specifications the coefficient on NotExemptit is large and positive. That is, not
exempt firms had higher productivity than exempt firms. This could be a reflection
17The fact that ?2 is negative in all the specifications might point towards the fact that the
government was choosing low productivity industries for de-licensing. This means that our coef-
ficients may still be biased due to industry-specific un-observables. However industry-year fixed
effects can capture these omitted industry-level factors and our results are robust to these effects
for the 1980s as well as for the full time span 1980-94 (Columns 1 and 2 of Table 3.6).
80
of the economies of scale enjoyed by larger firms.
In Appendix B.3 we show the results of the basic regression for Special in-
dustries. These were industries in which there was no distinction between exempt
and not exempt firms. This allows us to test an implication of our identification
strategy that there should be no variation in the performance of exempt and not
exempt firms due to de-licensing of special industries. We find that this is the case
in our data.
In Appendix B.4 we provide details of another specification test. One impli-
cation of our identification assumption is that if we choose arbitrary thresholds for
exemption from licensing, then we should not find any variation in the response of
firms to de-licensing around these arbitrary thresholds. That is, if the true threshold
matters then randomly chosen thresholds should not. We find that this is true in our
data. The significance level of the interaction coefficients when plotted against dis-
tance from the actual threshold takes a bell-shape. That is, the interaction between
de-licensing and the random threshold (both above and below the actual threshold)
is insignificant far from the actual threshold and rises in significance as we approach
the true threshold.
Our baseline results are robust to a variety of checks which are presented in
Table 3.6. One of the first things that we may be worried about is the possibility
of sample selection. In our main regressions we are excluding firms in Schedule IV
and V industries. It is possible that the Schedule IV and Schedule V industries
(which did not have exemptions from licensing for any firms) may have special
characteristics that may be correlated with size or with productivity. As we see
81
in Appendix B.3 (Column 2, Table B.4), the coefficient on the interaction of de-
licensing with exemption status for non-schedule IV and V industries is significant
and positive even when we include Schedule IV and V industries separately18.
In Table 3.6, Columns 1 and 2 we subject our results to a strict check. We
are not satisfied that industry-specific time trends capture all the industry-specific
heterogeneity that we may falsely attribute to reforms. Hence we include 2-digit
industry-year effects into our main regression as well as in the regression for the
years 1980-90. We find that magnitude of the interaction coefficient is lower than
the baseline specification but it is still quite large and significant at the 10% level.
For the period 1980-90, the coefficient on the interaction is positive and significant at
the 19% level. We also test our results using 4-digit industry effects interacted with
year effects. Note that this means that the coefficient on Dejt is not identified since
Dejt varies at the 4-digit industry level. However the interaction term Dejt ?Neit
is identified and the coefficient on it is 0.15 and is significant at the 1% level. There
is also a possibility that our results may be driven by some outliers. Column 3 of
the table shows the results of our main regression when we winsorize the dependent
variable by 1%.
When we control for industry-year effects (Table 3.6), we are worried that the
time paths of growth might be highly dissimilar across industries and we wish to
control for that. In Column 4 of Table 3.6 we subject our result to a stricter check.
We include industry-state-year effects. Our concern is that industries in different
18We also tried regressions including non-factories in the sample. Our results are robust to the
inclusion of these firms.
82
states may have different time paths of growth. This could be due to different growth
paths of infrastructure, financial markets, labor laws, wage to rental ratios etc in
different states. These factors may affect the incentives and the ability of firms lo-
cated in different states differently. If de-regulated industries are disproportionately
located in states that have higher growth rates of these variables then we maybe
falsely attributing this to the reforms (rather than to the growth of these variables).
We find that after controlling for three-way 3-digit industry, state, year effects the
average affected firm (with NeitDejt = 1) still has higher productivity than oth-
ers. Further note that in this specification the average impact of de-licensing (?1)
is positive. That is, after controlling for industry-state-year interactions all firms
in de-licensed industries performed better than firms in licensed industries. This
means that the total impact of de-licensing (?1 +?3) is quite large.
3.6.2 Controlling for firm-level un-observables
A very important source of bias in our co-efficient estimates may be due to reverse
causality. The main causal story that we want to tell is that the productivity of a firm
that was under licensing requirements (NotExemptit = 1) and was in an industry
that was de-licensed by the government will be affected by de-licensing. However,
it may be that a firm of a particular productivity may chose to be exempt from
licensing. That is, firms that have low productivity may have a systematic tendency
to stay exempt from licensing requirements since that keeps them protected from the
larger, licensed firms. Then, since low productivity firms are choosing to be ?non-
83
treated? firms, we will see a large effect of de-licensing even if it did not have a large
impact. That is, our estimates of the impact of de-licensing are biased upwards.
On the other hand, it is also possible that low productivity firms chose to
be under licensing even though there are constraints. This is because a license to
produce was permanent and entry into an industry was not governed by market
conditions. Thus there would be a high chance of a low productivity firm surviving
and even making super-normal profits without inducing entry from competitors. In
this case, treated firms in our sample are systematically low productivity and this
will bias our estimates downwards.
Another story we can think of is that firms with low managerial ability chose
to be under licensing since they know they will not be able to compete in a more
competitive environment. Again, if we are not able to control for these firm-level
effects then our estimates may suffer from omitted variable bias.
Ideally, we would like to verify that the results reported above are robust
to the inclusion of firm-level fixed effects. That is, we would require information
on firm identity and the ability to follow a firm over time. In our data we can
theoretically identify 200,000 observations over time (those belonging to the census
sector). However, we have not been provided with firm identity numbers by the
ASI. But we do have rich data on a variety of firm identifiers and we try to use
those data to construct a firm identity number.
We use information on 6 firm-level identifiers:
1. Four-digit industry of production;
84
2. State where firm is located;
3. Whether located in rural, urban or metropolitan area;
4. The year that the firm started production;
5. The ownership structure of the firm (government owned, wholly privately
owned or joint sector)
6. The area in which a firm is located.
There are other firm-level identifiers that could be used like the organization
structure of the firm (proprietorship, co-operative society, private corporation etc)
and a finer division of the ownership structure (whether owned by the state or
central government). However we do not use these since they might change over
time for a given firm. Thus, it is possible that a firms organization changes over
time from a joint family proprietorship to a private limited company or that a
wholly central government owned firm becomes a joint partnership with the state
government. However, it is unlikely that a government owned firm (either state or
central government) will change to a wholly private firm over time.
The six variables that we have chosen are available for 198,942 observations.
Note that our sample of 200,000 observations also includes firms that were closed
during the accounting year. We cannot use these firms in our estimation since output
and employment figures are not available however we include closed firms while
creating the firm-level identifier. This means that the same identifying number is
assigned to a firm that is in the data and producing in year t, is closed for production
85
in year t+ 1 but enters production again in year t+ 2.
Using these identifiers, we are able to uniquely identify approximately 33,207
firms. For the other firms, there are multiple identity numbers within each year
and we drop those firms. Further note that each of these firms does not exist every
year. That is, we have only surviving firms in our sample. Our total sample size is
140,562 including closed firms and 100,000 excluding closed firms.
We are still concerned whether we are capturing the same firm over time. It
may be that another firm with very similar characteristics existed in the same region
and in later years we are capturing that firm. To check for this we use two pieces of
information from our data set- the opening and closing stock of fixed capital owned
by the firm. Suppose that firm 1 ended year 1980 with Rs. 1 million of capital stock.
Then we check whether the firm identified as 1 in 1981 started with capital stock
close to Rs. 1 million. This cross-check supports our firm identifier.
Our main concern is that this may not be a random sample. So we provide
some descriptive statistics of our panel of firms as compared to all factories. Fig-
ure 3.4 shows the distribution of productivity in exempt and not exempt firms for
the pseudo-panel and the full sample firms. The black solid line represents unique
firms while the red dashed line represents the all firms. As we can see that the
distribution of productivity is very similar for both types of firms over exemption
status. Any difference between unique firms and all factories are constant over ex-
emption status and hence will get canceled out. For example, the sample of unique
firms has higher productivity for both types of firms-exempt and not exempt from
licensing requirements. In Figure 3.5 we investigate the distribution of assets (in
86
plant, machinery, land, building) for exempt firm over de-licensing reform in the
full sample and in the pseudo-panel. Within each graph the solid line represents
the year 1988 while the dashed line represents the year 1983. Comparing this graph
to Figure 3.1 (that was drawn for the full sample) we can see that the behavior of
exempt firms over licensing reform and over time is the same for the full sample as
for the pseudo-panel.
The results of the estimation of Equation 3.3 on uniquely identified firms are
presented in Column 1 of Table 3.7. We estimate a within-firm regression with year
dummies and standard errors clustered around 4 digit industry. The coefficient on
the interaction of exemption status with de-licensing is positive and significant at the
5% level. That is, firms that were treated with de-licensing reform (they were not
exempt from licensing and were in industries that were de-licensed) performed 5.7%
better on average than non-treated firms, conditional on firm-level factors (the total
impact of deregulation on not exempt firms is ?1 + ?3). In Column 2 of Table 3.7
we include 2-digit industry-year fixed effects to capture different growth paths of
different industries and we find that similar to our baseline results, the coefficient
on the interaction term is lower in magnitude. The average treated firm did 7.7%
better than a non-treated firm, even after controlling for firm-specific factors and
for industry-year interactions. In Column 3 of the table we present the results for
the year 1980-90. The coefficient on the interaction is 0.079 and is significant at
the 10% level. That is, after controlling for firm-specific factors, we find that the
reforms of the 1980s had a positive and significant impact on the firms that were
affected by these reforms.
87
3.7 The Relationship between Trade and Industrial Reforms
3.7.1 Does Deregulation Matter?
As mentioned earlier, one important feature of the Indian case is the interesting
chronology of reforms. In the 1980s, policy was geared towards de-regulating in-
dustry to reduce administrative burden as well as unnecessary checks on firm-level
decision making. There was little interest in trade reforms and the changes in
trade policy that occurred were primarily of the nature of streamlining of the sys-
tem, rationalization of administrative procedures etc. It was only in 1991, under
pressure from international organizations, that the first steps were taken towards
trade reforms. Following the industrial policy statement of August 1991, there were
across-the-board tariff cuts bring down the average un-weighted tariff to 60%19.
Given this chronology of events an important question that arises is what
happened to the industries that were deregulated in the 1980s post-trade reform?
In this section we will try to assess the impact of the 1991 changes in industrial and
trade policies on firm-level productivity.
In Column 1 of Table 3.8 we present a simple estimate of the post-1991 perfor-
mance of firms in industries that were de-licensed in the 1980s . We define D80jt =
1 if industry j was de-regulated in year s ? t of the 1980s, 0 otherwise. That is,
the D80jt variable is the same as the Dejt variable. We simply don?t include the
19The strength of pre-reform trade barriers and hence the scope of the reforms can be gauged
from the fact that according to Panagariya (2004), in 1990-91 the highest tariff rate stood at 355%,
the simple average of all tariffs at 113% and the import-weighted average of tariff rates at 87%.
88
industries that were deregulated in 1991 and 1993. Our estimation equation is
yijts = ?0 +?t +?j +?0D80jt +?1NotExemptit +?2Post91t (3.4)
+?3D80jNotExemptit +?4D80jNotExemptitPost91t
+?5D80jPost91t +?Xjts +?jts
The coefficients ?0, ?1 and ?3 have the same interpretations as in Equation 3.3.
The coefficient on NotExemptitD80jt is the impact of de-licensing on the firms
that were affected by it . The coefficient ?4 then measures impact of deregulation
(conducted in the 1980s) on the affected population of firms in the post-trade reform
period (Post-1991). Column 1 of Table 3.8 shows that this estimate is positive and
significant. That is, the average productivity of affected firms (not exempt from
licensing and in de-licensed industries) in the post-trade reform period was 8.4%
higher than that of firms that were not deregulated during the 1980s. To get a
sense of the impact of deregulation and trade relative to general macroeconomic
trends we also present the coefficients on the year dummies below Table 3.8. In
the second column of the table, we include the variable D80j ? Post91t. We are
worried that without this term, the coefficient on NotexemptitD80jPost91t might
be capturing differential trends in productivity in de-licensed industries, post 1991.
As Column 2 shows, the coefficient ?4 declines in magnitude after we control for this
and significant at the 16% level. We do not include an interaction of NotExemptit
and Post91t since size-based exemption of firms from licensing provisions was no
longer a government policy after the reforms of 1991.
In Column 3 of Table 3.8, we check the robustness of our results by running
89
within-firm regressions using our pseudo-panel. Note that even after controlling
for D80jPost91t, there is 13% higher productivity for firms that were deregulated
in the 1980s, in the post-1991 period. This confirms our main result that firms
that were affected by de-licensing during the 1980s performed better than other
firms in the post-1991 period. This points towards a complementarity relationship
between trade and industrial policy. That is, a rise in domestic competition raises
the marginal response to a trade reform episode.
To assess the immediate and medium-term impact of trade reforms, in Column
4 of Table 3.8 we split the post 1991 period into two subperiods. We define Trade1 =
1 if year is 1991 or 1992, 0 otherwise, and Trade2 = 1 if year is 1993 or 1994, 0
otherwise, and interact NotExemptitD80jt with both of these variables. As we can
see from Table 3.8, the immediate short-term impact of the trade reforms was higher
productivity for firms that were affected by de-licensing during the 1980s while the
medium term impact is insignificant. As we get more firm-level data for the post-
1991 period, we might be able to draw firmer conclusion about the short-, medium-
and long-term impact of these reforms.
3.7.2 Testing Predictions from Theory
The theoretical model in Chapter 2 postulates a two-step relationship between trade
and industrial policies and the innovation incentives of the firm. The first is the link
between trade policy and innovation incentives. The second is the link between
industrial policy (deregulation) and the response of innovation to trade reform.
90
The link between trade policy and innovation incentives is mediated by the
technological status of the industry. Specifically, trade liberalization encourages
innovation in industries that are more advanced technologically and discourages
innovation in other industries. The intuition for this result is that not-so-advanced
industries know that they will not be able to compete with foreign firms (assumed to
be very advanced) and would prefer to cut their losses by not investing in technology.
On the other hand, advanced industries realize that they have a good chance of
deterring entry by foreign competitors if they invest in technology and manage to
become productive enough to deter entry by foreign firms.
The second link is that between industrial policy and innovation. Our theory
postulates that there is a relationship between trade and industrial policies vis their
effect on a firm?s incentives to invest in innovation. That is, under certain condi-
tions20 an industry that faces more domestic competition will have a higher marginal
response to the higher threat of entry following a trade liberalization episode. A
technologically advanced industry will raise its investment in productivity in re-
sponse to a more competition from abroad, but this response will be higher when
he faces more competition at home.
This strategic complementarity relationship means that productivity will be
higher in industries that are initially more competitive (in the case of India, due to
20The condition is that there is a high degree of monopolistic distortion prior to industrial
deregulation. That is, the price-cost margin of the industry is above a certain level. This condition
will be true for industries that were subject to industrial licensing since by default, entry into the
industry was determined exogenously and was not a result of market forces.
91
industrial deregulation that removed barriers to entry into industry) and now face
a higher threat of entry from foreign competitors. Further given the relationship
between the technological nature of the industry (advanced or backward) and trade
reform, this positive relationship between industrial deregulation and trade reform
rises as the industry becomes more advanced i.e. improves in technology. This
translates into a prediction of higher productivity in industries that were deregulated
in the 1980s and had higher productivity (i.e. were more advanced), in the post-1991
period.
There are two points to note about our result. The first is that we do not
explicitly model trade. That is, the trade reform is modeled as a rise in the threat
of entry from foreign competitors and advanced industries react by competing while
backward industries by retreating. That is, the only impact of trade that we are
interested in comes through the resultant increase in competition. The second point
to notice is that in our case, industrial deregulation was followed by a rise in com-
petition due to a trade reform episode.
Thus, an interesting test of our predictions from the theory is the following.
Do industries that were affected by the first rise in competition during the 1980s
and were technologically advanced as a result of that, become more productive after
the second rise in competition during the 1990s?
We try a direct test of our theoretical prediction by analyzing the performance
of firms in industries that were deregulated in the 1980s, in the period after 1991.
Note that de-licensing in the 1980s occurred in the years 1984, 1985, 1986 and 1987.
So the period 1988-1990 can be used to measure the medium-term impact of de-
92
licensing in the absence of trade reform. The period 1991-94 can be used to measure
the immediate impact of the trade reform episode.
Now we include a variable to indicate the technological status of an industry.
We define the technological status of an industry based on its performance in the
years 1983 and 1988. We define
Initialj =
??
???
????
1 industry j has higher productivity in 1988 compared to 1983
0 otherwise
Our regression equation, which we estimate for the period 1988-94, is given
below
yijts = ?0 +?t +?j +?0InitialjPost91t +?1D80j +?2D80jPost91t (3.5)
+?3D80jIntialjPost91t +?Xijts +?ijts
Here yijts is the productivity of firm i industry j at time t, located in state s.
The coefficient on InitialjPost91t is a test of our prediction that technologically
advanced industries perform better after a trade reform episode. The coefficient on
D80jIntialjPost91t is a test of our prediction that the marginal response to trade
reform rises as the industry faces more competition domestically. Thus ?3 > 0 will
confirm our prediction that industries respond more aggressively to trade reform
if they face more competition domestically21. We include D80j to control for any
special features that these industries may have. We include D80jPost91t to control
21In our definition of the initial productivity status of the industry there is an implicit assumption
of spill-overs from deregulation. As mentioned earlier, only Not Exempt firms were under the ambit
of licensing and hence were affected by de-licensing. Since we define Initial using the average
productivity of an industry, both exempt and non-exempt firms are included in the calculation of
93
to allow for heterogenous response to the macroeconomic shock in the deregulated
industries.
In Table 3.10 we present the estimates of this equation for the period 1988-
94. In Column 1 we control for 2-digit industry-year effects (to control for time-
varying heterogeneity in industry growth paths), state effects and firm-level controls
like ownership, organization, location etc. These are within-industry regressions.
We use 4-digit industry effects and since D80j varies at 4-digit industry, ?1 is not
identified. However the coefficient of interest is identified. As Column 2 shows ?3
is positive and significant. Thus, industries that were de-regulated in the 1980s and
were technologically advanced did better in the period of high competition following
trade reform22.
An important point to note from Column 1 is that the coefficient on ?2 is
negative and insignificant. That is, industries that were deregulated during the
1980s on average performed the same or worse in the post-1991 period. However
the impact of D80j in the post-1991 period is ?2 +?3 ?Initialj. Thus an advanced
Initial. Thus, the implicit assumption is that there were spill-overs from the de-licensing of Not
Exempt firms on the productivity of Exempt firms. This could be because Exempt firms were
forced to deal with freer entry post-reforms, in all types of firms. This means that industries where
the Exempt firms also became more productive after de-licensing are more likely to have done well
between 1983 and 1988.
22In the context of our discussion of the definition of Initial, this result implies that industries
where even the Exempt firms became more productive after de-licensing, raising average industry
productivity, did really well after trade reforms. Thus, the impact of trade reforms is enhanced
when there are productivity spill-overs to firms that were not affected by de-licensing.
94
industry that was deregulated in the 1980s had 17.8% higher productivity on average
in the post-1991 period (since ?2 is not significant).
Note that ?0 is negative. That is, industries that were doing well during the
1980s performed worse after trade reforms. However, the total impact of trade
reforms is ?0 + ?3 for advanced industries that were deregulated during the 1980s
and this is positive. One possible reason for the negative impact of trade reform on
advanced industries could be that there is no notion of comparative advantage in
our model and estimation. It is possible that some industries may be technologically
advanced prior to trade reform but they collapse after trade reforms since they do
not have a comparative advantage in that product. As a caveat, in a differentiated
product model such as ours, some goods in all industries could be exported after
trade reform unless firm productivity falls too low and it can not overcome a trading
cost.
In Column 2 we present results for when we control for 3-digit industry rather
than 4-digit. The purpose of this is to be able to identify the coefficient on D80j.
This analysis bears out our basic results. The coefficient ?3 is positive confirming
the strategic complementarity relationship between trade and industrial policy. The
total impact of deregulation during the 1980s was to raise the post-1991 productivity
of those industries (?1 +?2) by 6.6%.
It is interesting to note that our results are very similar to the ones pre-
sented here when we try a very different measure of whether the firm is advanced
or not. Rather than comparing inter-temporal advance in the technical status of
the industry, we compare average productivity of each industry with the maximum
95
productivity in that industry in the year 1988. That is, we take into account not
only the highest productivity in an industry but also the dispersion of firms accord-
ing to productivity, within the industry. And our main result, that industries that
are more productive and were deregulated during the 1980s performed better after
the reforms of 1991 holds. That is, this very different measure of distance from the
technological frontier also provides evidence of the strategic complementarity rela-
tionship between trade and industrial policy. However here too we get the puzzling
result that industries that are closer to the frontier, perform worse after 199123.
Our preliminary and simple tests show that firms in industries that were de-
licensed in the 1980s and hence faced a more competitive, market-oriented envi-
ronment as well as lesser restrictions on their abilities during the 1980s, performed
better after a trade reform episode as compared to firms in industries that were
23A possible explanation for this result is a composition effect. Our measure of the technological
status of the industry takes in to account average industry productivity and compares it with
the maximum industry productivity. Suppose that firms are heterogenous in productivity within
an industry and there are fixed and variable costs to exporting (a Melitz (2003) set-up). After
trade reforms, firms with the lowest productivity exit (or supply only the domestic market) while
the highest productivity firms expand and their labor productivity rises. Keeping in mind our
definition of the frontier, in an industry where there are many productive firms, trade reform
could lead to a decline in recorded productivity since these firms lose sales to foreign varieties.
In an industry with a large dispersion of productivity, the least productive firms exit while the
most productive firms expand and record even higher output per worker. Thus industries that are
?advanced? see a decline in productivity while the ?backward? ones see a rise in productivity. This
point again reveals that integrating an explicit trade model into our model might yield interesting
and important predictions
96
not de-regulated in the 1980s. Thus the strategic complementarity relationship pos-
tulated by our theory seems to hold for the case of India. Future work that ties
together our theory with a trade model may shed more light on the our model and
predictions.
3.8 Conclusions and Future Work
In this chapter we use two unique data sets, the institutional features of Indian
policy and the interesting chronology of reforms in India to address two issues.
The first is whether industrial de-regulation that increases the level of competition
that a firm has to face leads affects firm-level productivity. We find the answer
to this is affirmative. Confirming our intuition as well as predictions from our
theoretical model we find that firms that were affected by de-licensing had higher
labor productivity than non-affected firms. Thus, lesser controls on ability as well
as more competition and a higher threat of potential entry spur firms to perform
better. We solve the problem of industry-level omitted variables that may bias
our estimates by using the fact that some firms were exempt from licensing. By
differencing over these two types of firms we are able to get consistent, unbiased
estimates of the impact of reforms.
The second issue that we address is whether there is a relationship between
industrial deregulation and trade reform. Given the chronology of reforms in India,
what happened to the productivity of the firms that were in industries deregulated
during the 1980s after the trade reforms of 1991? We find that firms in industries
97
that were de-licensed in the 1980s and had higher productivity as a result of that
tended to have higher productivity post-1991.
Our results are robust to the inclusion of a wide variety of firm- and industry-
level controls and fixed effects. One implication of using fixed-effect methods to
assess the impact of reforms is that we can test an implication of our identification
assumption-for the set of industries were exemption from licensing was not available
there should be no variation in the performance of firms due to de-licensing around
the threshold for licensing-and we find that this assumption holds. In another speci-
fication test, we check for the significance of randomly chosen thresholds for licensing
and find that only the official threshold matters with respect to differential perfor-
mance in exempt and not exempt firms. Further we find no evidence of bunching
of firms around the threshold for licensing and no evidence of firms choosing their
exemption status based on anticipations of de-licensing reform. We also construct
a panel of firms to test the robustness of our estimates to the inclusion of firm fixed
effects.
Our results have interesting policy implications. An important one is that
domestic competitive environment can be used to prepare firms in the economy
for trade reforms. Under competition from high-productivity foreign firms, domes-
tic firms that are not productive may want to cut their losses and not invest in
productivity-enhancing technology. However, a rise in the level of domestic compe-
tition and lesser controls on microeconomic decisions can spur these firms to make
investments in technology prior to facing competition from abroad and hence pre-
pare them for an even more competitive environment.
98
There are several important issues that remain to be answered about the
Indian reforms. The first is the debate about the unexplained rise in the total factor
productivity of Indian manufacturing prior to the reforms of 1991 (Delong (2001),
Panagariya (2001), Panagariya (2002), Panagariya (2004)). Our estimates of labor
productivity reveal that there was rise in productivity during the 1980s and that
this was the result of de-licensing. We can extend this study to other measures
of productivity and assess whether total factor productivity rose as a result of de-
licensing.
Another issue is that of the mechanism through which de-licensing affected
firm productivity. Our current estimates measure both the impact of removal of
controls on ability as well as incentives. It may be possible to delineate the role of
firm-level incentives to invest in technology as a result of more competition from the
role of firm-level ability to invest in technology as a result of de-regulation. Licensing
curbed the ability of the firm because many firm-level decisions (location, amount
produced etc) were decided by licensing authorities. We have detailed information on
these policies and analysis may shed more light on the issue of ability vs. incentives
as well as on the role of locational or output constraints.
We can test another prediction from our theoretical model that the produc-
tivity is affected by the level of competition in your own sector as well as the level
of competition in your upstream sector(Propositions 3 and 4, Section 2.3.3). The
intuition for this result is that a reduction in the monopolistic distortion in the
upstream sector raises the bargaining power of the downstream sector, allowing it
to recoup more of the surplus from its investment in technology. Thus, even if the
99
downstream industry is not deregulated there should be a rise in its productivity
if its principal upstream buyer is deregulated. Using input-output matrices for the
1980s and 1990s we can identify the principle buying sector for each industry and
test this prediction.
We also plan to assess the impact of a firm?s technological status (as opposed
to an industry?s technological status) on its response to more competition from
abroad. For this, we can use our pseudo-panel of firms to identify those firms that
saw an increase in their productivity between 1980 and 1990 and showing that they
experience an increase in productivity after 1991 greater than the ones that did not
see a rise in productivity between 1980 and 1990.
Finally, as mentioned earlier it might be very informative and interesting to
try to incorporate a trade model into our simple theoretical framework in order to
gain more insight in to the mechanisms through which trade can affect firm-level
productivity.
100
Tables
All Output Factory Output All Value Added Factory Value Added
1984 7.00 7.10 10.00 10.10
1985 18.30 20.30 27.70 29.00
1986 3.80 3.60 3.70 3.70
1987 26.50 27.90 32.70 32.50
1991 59.10 58.30 56.90 56.30
1992 0.00 0.00 0.00 0.00
1993 2.60 2.80 3.20 3.20
Table 3.1: Percentage of Output and Value Added De-licensed in each year. Source:
Annual Survey of Industries; ?Guidelines for Industry?, Ministry of Industry (var-
ious issues) and own calculations. Note: Factories are defined as enterprizes using
power in the production process and 50 or more workers or enterprisers not using
power in the production process and 100 or more workers. According to the act
under which licensing was promulgated in India, these factories were the ones under
licensing.
Year Inverse of Sampling Probability
1 2 3
1980 13,345 2,880 0
1981 13,746 2,910 0
1982 14,771 2,913 0
1983 13,883 2,319 0
1984 13,398 2,240 0
1985 13,785 2,439 0
1986 13,747 2,236 0
1987 11,080 2 2,786
1988 11,183 1 2,804
1989 11,195 186 2,880
1990 10,894 321 2,801
1991 11,202 353 2,752
1992 11,739 289 3,183
1993 11,915 312 3,413
1994 12,607 349 3,380
Table 3.2: Characteristics of the data
101
Variable Mean Std. Dev.
Not Exempt Firms (Observations=12825)
Ln (Output per Worker) 8.6 1.113734
Real Output (1993-94 Rs.) 1.25E+07 2.95E+07
Book value of Assets in 4.45E+08 1.98E+09
Plant, machinery, land and building (Rs.)
Employees 1814.088 3369.832
Exempt Firms (Observations=286567)
Ln (Output per worker) 7.22921 1.452706
Real Output (1993-94 Rs.) 591649.6 1333756
Book value of Assets in 6612464 1.53E+07
Plant, machinery, land and building (Rs.)
Employees 219.8095 683.028
Table 3.3: Summary Statistics for All Factories. Factories are defined as enterprizes
using power in the production process and 50 or more workers or enterprizes not
using power in the production process and 100 or more workers. According to the
act under which licensing was promulgated in India, these factories were the ones
under licensing. Not exempt firms are those that were under licensing, exempt firms
are those that were granted an exemption from licensing based on the size of their
assets.
Exempt Firms in Licensed Industries
Mean Median 75th Percentile
1980s 16.99 17.13 17.77
1980-83 26.26 29.73 23.81
1984-89 6.70 11.52 12.97
Exempt Firms in De-licensed Industries
Mean Median 75th Percentile
1980s 15.45 14.09 15.96
1980-83 20.04 16.96 19.39
1984-89 7.05 10.88 10.53
Table 3.4: Average Annual Rates of Growth of Assets(%). Schedule IV and V
industries are not included here. Assets refers to the book value of assets in plant,
machinery, land and building.
102
Note: For Tables 3.5 to 3.10, *** refers to significance at the 1% level, ** to
5% and * to 10% level. Controls are ownership, organization, location of the firm,
whether large firms were allowed to enter the industry, wage to rental ratio. The
dependant variable is log(Output per worker). Each observation is weighted by its
sampling probability. NotExemptit = 1 if firm was under licensing requirements,
Dejt = 1 if industry was de-licensed in year t. For all regressions standard errors
are clustered around 4-digit industry.
Baseline Baseline Industry-trend Industry-trend
1980-90 1980-90
NotExempt 0.593*** 0.587*** 0.607*** 0.598***
[0.100] [0.096] [0.096] [0.096]
De -0.055*** -0.093** -0.064*** -0.067**
[0.021] [0.042] [0.020] [0.033]
De*NotExempt 0.161** 0.148 0.132* 0.102
[0.078] [0.112] [0.073] [0.114]
Constant 6.252*** 6.428*** 6.187*** 6.412***
[0.150] [0.209] [0.134] [0.170]
Observations 172297 125898 172297 125898
R-squared 0.51 0.52 0.51 0.53
Industry-Trend No No Yes Yes
(3-digit)
Industry FE Yes Yes Yes Yes
(4-digit)
State FE Yes Yes Yes Yes
Year FE Yes Yes Yes Yes
Table 3.5: Baseline Results. These regressions do not include Schedule IV and
Schedule V industries.
103
Ind-Year Ind-Year Winsorised Ind-State
Effects 1980-90 -Year Effects
NotExempt 0.612*** 0.595*** 0.590*** 0.673***
[0.097] [0.098] [0.093] [0.022]
De -0.008 -0.015 -0.01 0.338***
[0.024] [0.029] [0.023] [0.037]
De*NotExempt 0.128* 0.128 0.113* 0.094***
[0.070] [0.109] [0.067] [0.032]
Constant 6.260*** 6.608*** 6.270*** 6.415***
[0.148] [0.220] [0.153] [0.037]
Observations 172297 115357 172297 172297
R-squared 0.51 0.53 0.52 0.62
Industry-Year Yes Yes Yes No
(2-digit)
Ind-State-Year Effects No No No Yes
Industry FE Industry FE Yes Yes Yes
(4-digit) (4-digit)
State FE State FE Yes Yes Yes
Year FE Year FE Yes Yes Yes
Table 3.6: Robustness Checks. These regressions do not include Schedule IV and V
industries.
Base Industry-Year 1980-90
Effects
1 2 3
NotExempt=1 0.126*** 0.141*** 0.109***
[0.032] [0.033] [0.036]
De=1 -0.048** -0.03 -0.0405
[0.024] [0.026] [0.034]
De*NotExempt 0.105** 0.077* 0.079*
[0.044] [0.040] [0.048]
Constant 7.323*** 7.370*** 6.81***
[0.053] [0.072] [0.071]
Observations 94714 94714 68348
R-squared 0.05 0.07 0.06
Number of group(id23) 33207 33207 26065
Industry Effects Yes Yes Yes
State Effects Yes Yes Yes
Year Effects Yes Yes Yes
Industry-Year Effects No Yes Yes
Table 3.7: Results for Pseudo-Panel of firms. These are within-firm regressions. In
columns 2 and 3 we include 2-digit industry effects interacted with year effects.
104
Baseline Baseline Pseudo-Panel Medium &
Robust Short run
NotExempt 0.644*** 0.644*** 0.150*** 0.644***
[0.094] [0.094] [0.027] [0.094]
D80 0.027 0.013 -0.082*** 0.027
[0.032] [0.030] [0.031] [0.032]
NE*D80 0.058 0.066 0.054 0.058
[0.112] [0.112] [0.035] [0.112]
NE*D80*Post91 0.084* 0.067? 0.137***
[0.046] [0.048] [0.045]
D80*Post91 0.038 -0.002
[0.035] [0.041]
NE*D80*Trade1 0.141***
[0.046]
NE*D80*Trade2 0.042
[0.058]
Post91 0.788*** 0.784*** 0.426***
[0.126] [0.128] [0.035]
Trade1 0.856***
[0.118]
Trade2 1.053***
[0.141]
Constant 6.259*** 6.258*** 7.210*** 6.258***
[0.148] [0.148] [0.020] [0.148]
Observations 172297 172297 95639 172297
0.51 0.51 0.04 0.51
2-digit Industry-
Year effects Yes Yes Yes Yes
Industry Effects 4-digit 4-digit 4-digit 4-digit
State Effects Yes Yes Yes Yes
Table 3.8: Impact of Trade Reforms. These regressions do not include Special
industries. ? represents significance at the 16% level.
105
Baseline Baseline Pseudo-Panel Medium &
Robust Short run
1981 0.126*** 0.126*** 0.050*** 0.126***
[0.035] [0.035] [0.016] [0.035]
1982 0.096** 0.096** 0.050*** 0.096**
[0.042] [0.042] [0.015] [0.042]
1983 0.194*** 0.194*** 0.076*** 0.194***
[0.050] [0.050] [0.023] [0.050]
1984 0.442*** 0.443*** 0.135*** 0.442***
[0.093] [0.092] [0.024] [0.093]
1985 0.562*** 0.564*** 0.233*** 0.562***
[0.098] [0.097] [0.034] [0.098]
1986 0.595*** 0.597*** 0.254*** 0.595***
[0.113] [0.111] [0.031] [0.113]
1987 0.582*** 0.585*** 0.250*** 0.582***
[0.106] [0.104] [0.029] [0.106]
1988 0.737*** 0.740*** 0.315*** 0.737***
[0.132] [0.131] [0.031] [0.132]
1989 0.784*** 0.787*** 0.405*** 0.784***
[0.116] [0.114] [0.032] [0.116]
1990 0.710*** 0.712*** 0.455*** 0.710***
[0.078] [0.077] [0.029] [0.078]
1991 0.068** 0.068** 0.029 0.068**
[0.032] [0.032] [0.026] [0.033]
1992 0.070** 0.054** -0.022 -0.068**
[0.031] [0.030] [0.020] [0.032]
1993 0.074** 0.074** -0.027 -0.190***
[0.030] [0.029] [0.017] [0.032]
1994 0.265*** 0.264*** 0.029 0.068**
[0.052] [0.052] [0.026] [0.034]
Table 3.9: Year Effects in Trade Regression
106
Within 4-digit industry Within 3-digit industry
Initial*Post91*D80 0.178** 0.511***
[0.074] [0.121]
Initial*Post91 -0.084** -0.161**
[0.042] [0.063]
D80 0.477***
[0.156]
D80*Post91 -0.113 -0.411***
[0.071] [0.110]
Constant 6.567*** 6.546***
[0.202] [0.199]
Observations 75476 75476
R-squared 0.46 0.44
2-digit Industry Year effects Yes Yes
Industry Effects 4-digit 3-digit
State Effects Yes Yes
Year Effects Yes Yes
Table 3.10: Impact of Trade Reforms on Advanced Industries. These regressions do
not include Special industries.
0
2.000e?07
4.000e?07
6.000e?07
0
10000000 20000000 30000000 40000000 50000000
0
10000000 20000000 30000000 40000000 50000000
Licensed De?Licensed
1983 1988
Density of Firms
Assets
Figure 3.1: Distribution of Assets of Exempt firms: Assets?Rs.50 million.
107
0
2.000e?08
4.000e?08
6.000e?08
8.000e?08
0
10000000 20000000 30000000 40000000 50000000
0
10000000 20000000 30000000 40000000 50000000
Licensed De?Licensed
1983 1988
Density of Firms
Assets
Figure 3.2: Distribution of Assets of Exempt firms: Rs.50 million?Assets>Rs.4.5
million.
0
2.000e?07
4.000e?07
6.000e?07
0
1000000 2000000 3000000 4000000 5000000
0
1000000 2000000 3000000 4000000 5000000
Licensed De?Licensed
1983 1988
Density of Firms
Assets
Figure 3.3: Distribution of Assets of Exempt firms: Assets?Rs.4.5 million.
108
0
.1
.2
.3
.4
0 5 10 15 0 5 10 15
Exempt Firms Not Exempt Firms
Unique Firms All Firms
Density of Firms
Assets
Figure 3.4: Distribution of Productivity in the Pseudo-panel. The solid line rep-
resents the distribution of unique firms-those that were assigned a firm id number.
The dashed line represents all firms.
0
1.000e?07
2.000e?07
3.000e?07
0
10000000 20000000 30000000 40000000 50000000
0
10000000 20000000 30000000 40000000 50000000
Licensed Industries De?licensed Industries
1983 1988
Density of Firms
Assets
Figure 3.5: Distribution of Assets of Exempt Firms in the Pseudo-panel. The dashed
line represents the distribution of unique firms in 1983, the solid line in 1988.
109
Appendix A
Appendices to Chapter 2
A.1 Proof of Equations 2.20 and 2.21
The equation for equilibrium investment by an advanced IG producer when the FG
sector is assumed to be perfectly competitive is given by Equation 2.19.
zA = ? 11?? (???)? ?11??LOt
bracketleftBig
(1 +g) ?1?? ?(1??)
bracketrightBig
= K (???)? ?11??LOt (A.1)
Here K ? ? 11??
bracketleftBig
(1 +g) ?1?? ?(1??)
bracketrightBig
. In the above equation, L0 refers to the
demand for labor by the FG sector that is given by the following equation.
LO =
?
?1 +? 11??
parenleftBigg1??
?w
parenrightBigg?
? ?11??A?
?
?
?1
?L (A.2)
We want to analyze the effect of a rise in ?, the price that the IG monopolist can
charge on his incentives to invest in productivity. That is, we wish to obtain the
total derivative of zA with respect to ? keeping in mind that ? is a function of the
parameter w, the wage rate in the economy. The relationship between ? and w is
given by the equation
? = 1?
?
?
parenleftBigg w?
1??
parenrightBigg??
+
parenleftBigg w?
1??
parenrightBigg1???
? (A.3)
Thus, any change in the wage rate will affect ? and hence zA.
110
For algebraic tractability as well as to get a meaningful expression for dzad? , we
make the simplifying assumption that both w and ?L change such that the allocation
of labor between the intermediate and final goods sector is unchanged. That is, we
allow w and ?L to change such that dL0 = 0. Thus, a change in the total endowment
of labor ?L occurs (holding L0 constant) and this leads to a change in the equilibrium
wage rate w (we explore the direction of this change below). In turn, a wage in w
results in a change in ? which affects the incentives of an firm to invest.
First we investigate the impact of a change in the labor endowment on the
wage rate. From Equation A.2, we see that
dL0 = ?
?L
D2 dD +
1
Dd
?L (A.4)
where D ?
parenleftbigg
1 +? 11??
parenleftBig1??
?w
parenrightBig?
? ?11??A?
parenrightbigg
. Setting DL0 = 0, we get
d?L = L0dD = (?1)L0? 11??A?? ?11??
parenleftBigg1??
w?
parenrightBigg?bracketleftBigg 1
(1??)?d?+
?
w?
?1
1??dw
bracketrightBigg
(A.5)
Hence the total derivative of ?L with respect to w is given by
d?L
dw = (?1)L0?
1
1??A??
?1
1??
parenleftBigg1??
w?
parenrightBigg?bracketleftBigg 1
(1??)?
d?
dw +
?
w?
?1
1??
bracketrightBigg
(A.6)
Using the relationship between ? and w given by Equation A.3, we get the
following equation for the total derivative of ? with respect to the wage rate w.
d?
dw =
1
?
?
w
parenleftBigg1??
w?
parenrightBigg?
(w?1) > 0 (A.7)
Plugging this into Equation A.6, we get
111
d?L
dw = (?1)L0?
1
1??A??
?1
1??
parenleftBigg1??
w?
parenrightBigg?bracketleftBigg 1
(1??)?
?
w
parenleftBigg1??
w?
parenrightBigg
+ ?w? ?11??
bracketrightBigg
< 0 (A.8)
This means that when the total endowment of labor ?L rises such that there
is no impact on the distribution of labor between the intermediate and final goods
sector then wages in the economy will fall. Intuitively this makes sense? a rise in
the supply of labor reduces wages in the economy.
Also note that from Equation A.7, a rise in the wage rate in the economy
leads to a rise in the price of the other input as well. That is, ? rises when w
rises (as a result of the change in the labor endowment). Intuitively, labor and the
intermediate goods varieties are imperfect substitutes in the final goods production
function. Thus when labor becomes more costly, the final goods producer tends
to reduce his demand for labor and raise his demand for the intermediate goods
varieties. This leads to an excess demand for intermediate varieties, leading to a
rise in their price. Another way to think about the positive relationship between
w and ? is that a rise in w contributes to rising marginal costs ? in the IG sector.
Given that each IG producer faces a constant elasticity of demand and constant
marginal costs, he will charge a constant mark-up over costs ???? = 11??. This in
turn means that when marginal costs rise as a result of a rise in w, in order to
maintain a constant mark-up the IG monopolist will raise his price ?.
The basic equation for investment in technology (Equation A.1) tell us that
the total derivative of zA with respect to w is
112
dzA
dw = K
?
?(???)? ?11?? dLOt
dw +L0t
d(???)? ?11??
dw
?
?
= KLOt
bracketleftBigg
? ?11??(d?dw ? d?dw)? (???)1?? ?( ?11??)?1 d?dw
bracketrightBigg
(A.9)
In step one, the first expression inside the square brackets is zero because of our
assumption that the distribution of labor between the two sectors remains unchanged
as a result of a change in w and ?L. That is, dL0 = 0. Now we use the relationship
between ? and ? given by ? = ??1? to get d? = ?d?. Substituting into the above
expression we get
dzA
dw = KLOt?
?1
1??
bracketleftBigg
(1??)d?dw ? (???)(1??)? d?dw
bracketrightBigg
= KLOt? ?11??
bracketleftBigg((1??)2 ?1)?+ ?
(1??)?
bracketrightBigg d?
dw (A.10)
In order to decide the sign of this expression, we use the fact that d?dw > 0 (from
Equation A.7) and that 0 < ? < 1. This means that ((1??)2 ?1) < 1 and hence,
dzA
dw > 0 ?
dzA
d? > 0 if ? > (1?(1??)
2)? ? ? ? ?? (A.11)
And the condition specified above holds when ? ? ??1. That is, at the profit
maximizing price of ? = 1??, the investment response of an intermediate goods firm
1To see this, assume that ? = ??. Then the condition ? > (1 ? (1 ? ?)2)? becomes ?? >
(1?(1??)2)? ? 1?(1??)2 > ? ? 1?(1??)2 ?? > 0. The last condition is true for ? > 0.
Thus, our initial assumption holds.
113
to a change in the price that it can charge (?) is positive. That is, a fall in market
power of the intermediate goods firm (as a result of a change in w and ?L) will lead to
a decline in the incentives of the firm to invest in productivity enhancing technology.
Note that to get the result in Equation A.11, we use the result in Equation A.7 that
a rise in wages (as a result of a change in ?L) will lead to a rise in ?.
In order to see the result in Equation 2.21, note that from Equation A.1,
sign(d2zAd?d?) = sign(dKd? ) ? sign(dzAd? ). The definition of K for an advanced IG firm
is K ? ? 11??
bracketleftBig
(1 +g) ?1?? ?(1??)
bracketrightBig
. Hence, sign(dKd? ) > 0 and sign(d2zAd?d?) > 0 for
an advanced firm. For a backward firm K ? ? 11??
bracketleftBig
(1??)(1 +g) ?1??
bracketrightBig
. Hence,
sign(dKd? ) < 0 and sign(d2zAd?d?) < 0
A.2 Cost Minimization Exercise of FG Monopolist
The final goods producer solves the cost maximization problem given below.
C(p(v),y) = MinLot,x(v),v?(0,1)wtLot +
integraldisplay 1
0
p(v)x(v)dv (A.12)
subject to yt ? 1?L0t1??
bracketleftBigintegraltext1
0 At(v)
?xt(v)?dv
bracketrightBig
.
This yields first order conditions
w
p(v) =
parenleftbigg1??
?
parenrightbigg(integraltext1
0 At(v)
?xt(v)?dv)
LoA(v)? x(v)
1?? (A.13)
p(v)
p(v?) =
parenleftBiggA(v)
A(v?)
parenrightBigg?parenleftBiggx(v?)
x(v)
parenrightBigg1??
(A.14)
Using the second set of first order conditions, we can get x(v?) as a function
of x(v)
114
x(v?) =
parenleftBiggA(v?)
A(v)
parenrightBigg ?
1??
parenleftBiggp(v)
p(v?)
parenrightBigg 1
1??
x(v) for all v? (A.15)
Using the above expression we can solve for (integraltext10 At(v)?xt(v)?dv) in terms of x(v),
integraldisplay 1
0
At(v)?xt(v)?dv = x(v)
?p(v) ?1??
A(v) ??1??
bracketleftbiggintegraldisplay 1
0
A(v) ?1??p(v) ??1??dv
bracketrightbigg
(A.16)
Using this expression in Equation A.13 we get the derived demand for variety v as
x(v) = ?wLot1??
?
?A(v)
?
1??
p(v) 11??
?
? ?A?1 (A.17)
where ?A ?integraltext10 A(v) ?1??p(v) ??1??dv is the price-weighted average productivity in the IG
sector. Note that the elasticity of demand is constant and hence a profit maximizing
monopolist producer in the intermediate goods sector will charge a constant mark-
up over costs. That is, p(v) = ? for all varieties v. Plugging this into the derived
demand function, we get
x(v) = ?wLot1??
?
?A(v)
?
1??
?A?
?
? (A.18)
where A? ?integraltext10 A(v) ?1??dv is the average quality in the intermediate goods sector.
Now we can use the above expression in the production function to get the de-
rived demand for labor by the final goods sector as a function of the output of the fi-
nalgoods sector. Finalgoods sectoroutputisgiven byy = Lo1??integraltext10 At(v)?xt(v)?dv =
bracketleftBig w?
(1??)?
bracketrightBig?
A?1??Lo. And so labor demand is Lo =
bracketleftBig(1??)?
w?
bracketrightBig? y
A?1??. Now we can sub-
stitute for Lo in Equation A.18 to get an expression for derived demand for the
115
intermediate good as a function of output y.
x(v) =
bracketleftbigg w?
1??
bracketrightbigg1???
? A(v)
?
1??y
?1??A?2??
?
? (A.19)
So the cost function for the final goods monopolist is
C(p(v),y) = MinLot,x(v),v?(0,1)wtLot +
integraldisplay 1
0
p(v)x(v)dv (A.20)
=
bracketleftBigg
w
parenleftbigg1??
w?
parenrightbigg?
+
parenleftbigg1??
w?
parenrightbigg??1bracketrightBigg
??A???1y
Note that the monopolist has constant marginal cost of production. His cost
rises as ?, the price of intermediates, rises and falls as the average quality of the
intermediates A? rises.
A.3 Proof of Propositions 1 and 2
Now we come to the case where the final goods sector is a monopoly. Now
zA = [(1 +g)
?
1?? ?(1??)]
A?2?? (???)?
??1
parenleftbigg w?
1??
parenrightbigg1??
ym
= C (???)???1
parenleftbigg w?
1??
parenrightbigg1??
ym (A.21)
Here C ? [(1+g)
?1???(1??)]
A?2?? and ym refers to the monopoly output of the final goods
sector firm and is given by the equation below.
ym = 12b
bracketleftbigg
a? M?
?
A?1??
bracketrightbigg
(A.22)
Here M is the marginal cost of production faced by the FG sector firm. This is
given by the expression
M ? w
parenleftbigg1??
w?
parenrightbigg?
+
parenleftbigg1??
w?
parenrightbigg??1
116
The total derivative of zA (Equation A.21) is given by
dzA =
bracketleftBigg?z
A
?w dw +
?zA
?? d?+
?zA
?? d? +
?zA
?ymdym
bracketrightBigg
(A.23)
The first term in Equation A.23 represents the direct impact of a rise in w on the
investment incentives of an IG firm and is given by
?zA
?w dw = C
parenleftbigg w?
1??
parenrightbigg??
?(???)???1dw > 0 (A.24)
That is, the direct impact of a rise in the wage rate on investment in the IG sector
is positive. This is because a rise in the wage rate reduces the demand for labor in
the FG sector and props up the demand for intermediate goods.
The next two terms in Equation A.23 represent the indirect impact of a rise
in w on z via its impact on the price ? and cost ? in the IG sector.
?zA
?? d?+
?zA
?? d? = C
parenleftbigg w?
1??
parenrightbigg1??
ym
braceleftBig
???1(d??d?)?(1??)???2(???)d?
bracerightBig
= C
parenleftbigg w?
1??
parenrightbigg1??
ym(1??)???1
braceleftBigg
d?? (???)?
bracerightBigg
d?
( using d? = ?d?)
= C
parenleftbigg w?
1??
parenrightbigg1??
ym(1??)???2?d? > 0 (A.25)
The above expression tells us that the indirect response of investment zA to a change
in w via changes in ? and ? is positive. A rise in ? means a rise in costs for the
intermediate goods sector firm and this has a negative impact on investment in
productivity. However, an increase in ? ameliorates this negative impact since a
rise in ? implies that the IG firm can appropriate more surplus from the FG sector
by charging a higher price.
117
Thus, when we substitute these expressions into Equation A.23 we get
dzA = C
parenleftbigg w?
1??
parenrightbigg1??
ym(1??)???2?d?+C
parenleftbigg w?
1??
parenrightbigg??
?(???)???1dw
+ C
parenleftbigg w?
1??
parenrightbigg1??
(???)???1dym (A.26)
Hence,
C?1
parenleftbigg w?
1??
parenrightbigg??1
?1??dzA = (1??)?ym? d?+ (1??)(???)ymw dw
+ (???)dym
C?1
parenleftbigg w?
1??
parenrightbigg??1
?1??dzAdw = (1??)?ym? d?dw + (1??)(???)ymw
+ (???)dymdw
(A.27)
Now we use Equation A.22 to solve for dym in Equation A.27.
dym = ??M?
??1
A?1?? d?+
???
A?1??dM
= ??M?
??1
A?1?? d?+
???
A?1??
parenleftbigg1??
w?
parenrightbigg?
dw
= ??(a?2bym)? d?+ ?(a?2bym)M
parenleftbigg1??
w?
parenrightbigg?
dw < 0
= a?? d?+
parenleftbigg1??
w?
parenrightbigg? a
Mdw?2bym
bracketleftBigg?
?d?+
a
M
parenleftbigg1??
w?
parenrightbigg?
dw
bracketrightBigg
(A.28)
< 0
In order to go from step one to two of the derivation above we use the definition
of M to get that dM =
parenleftBig1??
w?
parenrightBig?
dw. From the second to the third step we use the
expression ym = 12b
bracketleftBig
a? M??A?1??
bracketrightBig
(Equation A.22) to get that M??A?1?? = a?2bym. Thus,
the indirect impact of a rise in w on investment zA via output of the FG monopolist
ym is negative. This is because an increase in the wage rate increases costs for the
118
FG firm and it reacts by lowering output. Costs increase because the FG sector
employs labor (direct impact given by dM > 0) and also because d?dw > 0. That is,
a rise in wages leads to a rise in the prices of the intermediate goods as well.
Now we substitute this expression in for dym into Equation A.27. Thus, we
infer that the total impact of a rise in wages on invest in productivity in the IG sector
is negative when right hand side of Equation A.27 is negative. That condition is
given by
(???)|dymdw | > (1??)?ym? d?dw + (1??)(???)ymw (A.29)
Using Equation A.29, we can simplify this inequality in terms of ym and we
get the result that
dzA
d?
??
???
????
< 0 if y < ?y
> 0 else
(A.30)
Here the threshold level ?y is given by
?y ?
bracketleftBiga?
?
d?
dw +
a
M
parenleftBig1??
w?
parenrightBig?bracketrightBig
bracketleftBig2b?
?
d?
dw +
2b
M
parenleftBig1??
w?
parenrightBig?
+ 1?????
parenleftBig?
?
d?
dw +
???
w
parenrightBigbracketrightBig?1 (A.31)
Here d?dw is given by Equation A.7 as 1? ?w
parenleftBig1??
w?
parenrightBig?
(w?1) > 0.
Note that from Equation A.22, the condition ym < ?y translates into a condition
? > ?? since equilibrium FG output is negatively related to the price of intermediate
goods.
For the proof of Proposition 2, note from Equation A.21 that sign(d2zAd?d?) =
sign(dCd?) ? sign(dzAd? ). As shown above, sign(dzAd? ) < 0 for ? > ??. The definition
of C for an advanced IG firm is C ? [(1+g)
?1???(1??)]
A?2?? . Hence, sign(
dC
d?) > 0 and
119
thus, sign(d2zAd?d?) < 0 for ? > ??.The definition of C for a backward IG firm is
C ? [(1??)(1+g)
?1??]
A?2?? . Hence, sign(
dC
d?) < 0 and thus, sign(
d2zA
d?d?) > 0 for ? > ??.
A.4 Proofs of Propositions 3 and 4
Equations 2.28 and 2.29 give the expression for investment in technology done by
an IG firm. Suppose there is an increase in the equilibrium output of the FG sectro
ym (due to an change in parameters a or b). Then
zB = (???)???1
parenleftbigg w?
1??
parenrightbigg1?? (1??)(1 +g) ?1??
A?2?? ym
?
?zB
?ym = (???)?
??1
parenleftbigg w?
1??
parenrightbigg1?? (1??)(1 +g) ?1??
A?2?? (A.32)
> 0
And hence,
?2zB
???ym = (???)?
??1
parenleftbigg w?
1??
parenrightbigg1?? (?1)(1 +g) ?1??
A?2?? (A.33)
< 0
Similarly
zA = (???)???1
parenleftbigg w?
1??
parenrightbigg1?? [(1 +g) ?1?? ?(1??)]
A?2?? ym
?
?zA
?ym = (???)?
??1
parenleftbigg w?
1??
parenrightbigg1?? [(1 +g) ?1?? ?(1??)]
A?2?? (A.34)
> 0
And hence,
120
?2zA
???ym = (???)?
??1
parenleftbigg w?
1??
parenrightbigg1?? 1
A?2?? (A.35)
> 0
A.5 Regulated Monopolist in the Final Goods Sector
In this section we analyze the implications of modeling the final goods sector?s
market structure as that of a regulated monopolist. The form of the regulation
is that the monopolist needs to produce on or below a defined level yC. This is
consistent with the provisions of the license regime in India whereby every license
had an output limit printed on it and it was illegal to produce above this specified
limit.
From Appendix A.2 we know that the cost function for the unconstrained
monopolist is given by
C(p(v),y) = MinLot,x(v),v?(0,1)wtLot +
integraldisplay 1
0
p(v)x(v)dv (A.36)
=
bracketleftBigg
w
parenleftbigg1??
w?
parenrightbigg?
+
parenleftbigg1??
w?
parenrightbigg??1bracketrightBigg
??A???1y
That is, total cost is a linear function of the output of the firm y(.). In other words,
marginal cost is constant in output but is strictly increasing in ? and w, the price
of the inputs as shown in the equation below.
MC(w,?,A?) = ?C(p(v),y)?y =
bracketleftBigg
w
parenleftbigg1??
w?
parenrightbigg?
+
parenleftbigg1??
w?
parenrightbigg??1bracketrightBigg
??A???1 (A.37)
= W(w)??A???1
The marginal revenue for this monopolist is given by
MR(y(.)) = a?2by(.) (A.38)
121
The optimal supply function for the constrained monopolist is a function of
marginal cost. A constrained monopolist will produce the threshold level yC up till
the point that MC(?,w,A?) <= MR(yC). That is, at the given level of input prices
(? and w), the monopolist earns a positive profit. Now suppose that the level of ?
is such that MC(?,w,A?) > MR(yC) then the monopolist will produce according
to the schedule given by equation 2.22
ym = 12b
bracketleftbigg
a? M?
?
A?1??
bracketrightbigg
(A.39)
where M ? w
parenleftBig1??
w?
parenrightBig?
+
parenleftBig1??
w?
parenrightBig??1
.
That is, for lower levels of marginal cost (and hence, ?) the monopolist can
get as close as possible to his optimal profit maximizing output ym by producing
the constrained output level yC. Once marginal costs cross a threshold level, the
optimal profit maximizing output level shifts. Since the constraint is an upper bound
on output the monopolist can always produce a lesser amount if it maximizes his
profits. In Figure A.1 we show the profit maximization exercise of the constrained
monopolist.
The output level yM is the optimal output for an unconstrained monopolist
when the unconstrained firm faces a marginal cost curve MC(?1) and yC is the max-
imum output decided by the government. Given the constraint, the firm produces
yC since it is making a profit on each additional unit at that point. But suppose
that the marginal cost firm facing the firm is MC(?2) which is above MC(?C)
(the level of marginal cost that makes yC the optimal output) then the constrained
monopolist maximizes profits by producing at yM? which is on the supply curve of
122
the unconstrained monopolist. That is, the supply schedule of the constrained final
goods monopolist is given by
yM? =
??
???
????
yC if MC(?) ? MR(yC) or ? ? ?C
yM if MC(?) > MR(yC) or ? > ?C
(A.40)
Here we use the fact that marginal cost is strictly increasing in ?. This means
that the response of the FG sector output (and hence the IG sector investment in
technology) to liberalization of entry restrictions will be different depending on the
initial level of monopolistic distortion in the IG sector.
Price
yC yMyM?
MC($1)
MC($2)
MC($C)
Quantity 
Regulated Monopolist
MR AR
Figure A.1: Supply curve of Constrained Monopolist.
123
Now using Equations 2.28, 2.29 and 2.30,
?zAd
?? =
??
???
????
parenleftBig w?
1??
parenrightBig1?? [(1+g) ?1???(1??)]
A?2??
bracketleftBigyM(??+(1??)?)
?2?? +
???
?1??
?ym
??
bracketrightBig
if ? > ?C
parenleftBig w?
1??
parenrightBig1?? [(1+g) ?1???(1??)]
A?2??
bracketleftBigyC(??+(1??)?)
?2??
bracketrightBig
< 0 if ? ? ?C
(A.41)
Here ?C is the level of ? that makes yC (the government defined output level)
the optimal choice for the monopolist. Suppose that the initial level of monopolistic
distortion in the IG sector is not very high (i.e., a low value for ?). Then the FG
monopolist is on the inelastic portion of his supply curve. In that case, ?yM?? is zero
and only the Schumpeterian force operates on the IG firms. A rise in domestic
competition (modeled as a fall in ?) will lead each producer of intermediate goods
varieties to reduce their investment in quality-enhancing technology. However, if the
initial level of monopolistic distortion in the IG sector is very high (i.e., a high level
of ?) then it is possible (from Equation 2.30) to have a case when a fall in entry
barriers in to the IG sector will force the IG firms to invest more in technology.
This result holds for Proposition 2 as well. That is, for high initial levels of ?,
industrial and trade policies are strategic complements. That is, more competition
in the domestic market may raise the marginal response of domestic firms to a
trade liberalization episode even in the case when the FG sector is constrained with
respect to output2.
2A point to note here is that the condition for the bi-lateral monopoly effect to outweigh the
Schumpeterian effect in Equation 2.30, the condition is ? > ?? where the threshold level is given by
the implicit equation derived when we plug in Equation A.31 into Equation A.22. We would like
to compare ?? and ?C in order to get more insight. However, the solution to this implicit equation
is hard to obtain.
124
Appendix B
Appendices to Chapter 3
B.1 Background on Industrial Licensing in India
B.1.1 The Procedures
After independence from the British in 1948, India adopted a mixed economy frame-
work. The main pillars of this framework were a major role of the public sector in
economic development and control of private industry. These principles were cod-
ified in the Industries (Development and Regulation) Act of 1951 (referred to as
IDRA from now on). Under this act, the central government was given the right
to issue licenses to firms for entering or for continuing production in manufacturing
and certain basic industries were reserved for the public sector. That is, no private
firm could produce in those industries.
The main aim of licensing was to direct investment of ?precious? capital into
the desired industries. Policy-makers at that time felt that private industry could
not be trusted to invest in strategic and/or important industries like machine tools,
steel, cement, chemicals etc and would probably invest in quick-return industries
like consumer goods. Over time the licensing regime was used to fulfill a variety
of policy purposes like encouragement to small-scale firms, control of large firms
to prevent concentration of economic power in a few hands and to promote even
125
growth in all regions of the country.
A license was a document that permitted a firm to continue/begin production
in an industry. It was issued by the Ministry of Industry in New Delhi. Under the
IDRA, 1951 all factories (defined as enterprizes that did not use power but employed
more than 100 workers or enterprizes that used power and employed more than
50 workers) that were already operating or wished to operate in a specified list of
industries were required by the government to obtain a license. The point of licensing
was to direct investment into desirable directions and hence the Act specified a list
of important industries in which licensing would exist. These industries are referred
to as ?First Schedule? industries.
All applications for license were debated upon by the LicensingCommittee con-
sisting of officers from the administrative ministry (Industry), the Planning Commis-
sion and representatives of other government departments and the Director General
of Technical Development. This was the nodal body for technical recommenda-
tions. It gave recommendations to the licensing committee on crucial aspects of
the project. These recommendations included the optimal size of the project, the
technology to be used in the project, the amount of raw materials-domestic and
imported-required etc. The DGTD also granted annual allotments of raw materials
once the license was granted. The licensing committee hearing was at the end a
long chain of clearances that a firm had to obtain.
If the project was approved by the Committee then the firm was granted a
?Letter of Intent? valid for one year. If the firm was able to show sufficient progress
in the implementation of the project then it was issued a license. If production com-
126
menced within two years of the issue of the license then the license was permanent.
Otherwise the license was revoked.
Conversations with officials who were on the licensing committee during the
1980s revealed that the most important concern for the licensing committee while
debating a particular case was the demand-supply situation of the good. The gov-
ernment maintained highly detailed records of the exact production of the good
that was already taking place (all registered/licensed units were required to submit
detailed monthly production reports) and so had information on the supply of the
good. Information about capacity utilization of existing plants and demand pro-
jections from the Planning Commission were used to compute the demand side of
the equation. The applicant was also required to give demand projections for the
product. If it was felt that there was enough existing capacity to satisfy demand
then the application was rejected irrespective of the quality of the proposed good and
the nature and productivity of the technology that was proposed to be used. That
is, the new project was not assessed on the merit of its efficiency, productivity or
quality.
Another important facet was the type of the good. There was a disdain for vari-
ety in policy-makers of the time. Competition was thought to be wasteful and Nehru
once remarked ?Why do we need nineteen brands of toothpaste?? as reported by
Das (2000)1. This was especially true for ?luxury goods?-those which were deemed
to be unnecessary and could not be afforded by the masses-like air conditioners,
cosmetics etc. Another important consideration was import and foreign exchange
1Page 153.
127
requirements. A large number of applications were rejected because they required
?too? much foreign exchange.
B.1.2 Scope of Licensing
The scope of a license was fairly broad especially from the late 1960s onwards. The
conditions that a license specified were many. It specified the amount of output that
a firm could produce. It was conditional on the proposed location of the project.
Permission would be required to change location. The exact nature of the item to
be produced was also specified and the firm needed to take permission or another
license to change its product mix. Even the kind of technology and inputs that the
firm could use in production (though not specified on the license) was determined
by the Licensing Committee and the DGTD. This was because the most crucial raw
materials (steel, cement, fuel etc) were controlled by the government and the firm
needed to get annual allotments of these for production. Further a whole separate
license was required for any machine, equipment , components or raw materials that
needed to be imported. Even for that, the DGTD and the licensing committee made
recommendations to the Capital Goods Licensing Committee about the nature and
quantity of imports to be permitted.
According to Marathe (1989)2, the initial purpose of licensing was not to take
over decision-making from the firm. It was basically meant to guide investment.
However, the precarious foreign exchange position in the 1960s meant that imports
and hence foreign exchange needed to be controlled stringently. In order to obtain a
2Pages 85-86.
128
license to import raw materials or components, it became necessary for the firm to
specify in much more detail the quantity and the exact nature of the good it would
produce (in an example from Marathe (1989), the firm would need to specify that it
planned to produce multi-spindle lathes instead of simply machine tools). This was
because import licenses would be granted only for the exact components needed for
lathes. Capacity constraints also started to be important in the licensing process
because of the need to restrict imports. These details were primarily administrative
in nature and had no business on a legal document like a license but over time,
administrative convenience prevailed.
As the licensing mechanism became a tool for dealing with problems of ad-
ministrative allocation of foreign exchange, the interpretation of various clauses of
the IDRA became more and more strict and rigid. Even though the act itself took
a lenient view of ?substantial expansion? of production (under this clause the firm
needs to take permission to substantially expand production), gradually capacity
limits started being specified on licenses and over time, maximum capacity became
equated with maximum production permissible. Marathe (1989) attributes both
foreign exchange constraints as well as the control system feeding off of itself for the
fact that by the 1970s,
?.....the judgement of the Government and Planners on questions such as the
size, the nature of the equipment, the process and sometimes even the actual physical
location of a unit often prevailed over the judgment of the entrepreneur concerned.?3.
3Marathe (1989), page 88
129
B.1.3 Implementation of the License Raj
An important point that arises given the strict nature of the conditions that a
firm was obligated to follow is whether the firm had any incentives to follow those
conditions. Conversations with government of India officials who had been members
of the licensing committee in the 1970s and 1980s as well as officers of the DGTD
reveal that surprisingly, there were very little direct checks on the factory to see
whether the conditions were being satisfied or not. That is, at first glance a license
was a toothless piece of paper and there was no mechanism to make the entrepreneur
follow its conditions. As a reminder, a license restricted entry into an industry, the
amount produced by the firm, the product mix of the firm and sometimes, the
location of the factory.
The main way in which licensing requirements were implemented were by an
even more potent force than physical inspections. This force was the fact that all
essential raw materials for production- steel, cement, coal, fuel, furnace oil, railway
wagon movements, licenses to import equipment and raw materials etc- were not
freely available in the market. Each and every firm was allotted a certain amount of
these inputs each year based on the output limits specified on their license. Thus, it
was very difficult for the entrepreneur to produce over the limit on his license since
basic raw materials were allotted to him based on the licensed amount. Further
licensing officials also reveal that during the actual licensing process there was almost
no verification of the details provided by the entrepreneur. If he stated that his plant
was located in district A then this was taken as given. However once the license
130
was granted and the entrepreneur was petitioning the DGTD to allot him his quota
of raw materials for the year, each and every aspect of the project was thoroughly
scrutinized to check whether conditions were satisfied. Further any direct assistance
from the government (in the form of inputs, credit etc) legally obligated the firm
to send detailed monthly production reports to the DGTD. These reports were also
scrutinized for discrepancies.
Even though production without inputs allotted by the government would be
difficult, it was not impossible since there was a black market for these inputs. This
raises the possibility of entrepreneurs obtaining a license and then never going back
to the DGTD for allotment of inputs. This way they could locate anywhere and
produce whatever they wanted. Government officials were unwilling to put a number
on the percentage of license-holders that were forced to come back to the DGTD
for inputs and other permissions required for production. But when pressed, they
revealed that 95% of license holders who wanted to commence production within
two years of issue had to come to the DGTD for some allotment or the other and
would have to pass a highly rigorous investigation of their project.
Officers gave an example of their power over the license-holder when questioned
about the possibility that a lot of license-holders could have operated under the radar
and flouted the license conditions. Suppose an entrepreneur was issued a license that
specified that he was allowed to produce 10,000 units of a good per year. He would
need cement to build the plant where production would take place and he would
need equipment. The allotments of cement and equipment that were made by the
government were based on 10,000 units. So the factory would be so small that it
131
could not accommodate another machine even if the entrepreneur was able to obtain
it from the black market.
A different perspective on this issue is available from several government of
India reports. It was starting to become obvious in the 1960s itself that the licensing
regime was not able to fulfill its primary objective-to direct investment in desired
directions. Several committees and study groups were established during the 1960s
and 1970s that studied the impact and performance of the licensing regime (Hazari
(1967), Dutt (1969), Ramakrishna (1978) etc) also concluded that there was little
follow-up of licenses to ensure that the projects were taking off. For example the
Dutt Committee Report concluded that the licensing system had failed to prevent
growth of capacity in less essential industries and could not be expected to ensure
the creation of capacity in the more essential ones.
The Study Group on Industrial Regulation and Procedures under G.V. Ra-
makrishna (Ramakrishna (1978) in its report stated that ?installed capacity was in
several cases more than targeted capacity and in many more cases significantly lower
than the targeted capacity? 4. It also went on to observe that
?....the penal provisions of the Act have also not been very effective in en-
abling the enforcement of the conditions of industrial licenses. In many cases where
actual production exceeds the licensed capacity by several times, it has not been pos-
sible effectively to use these provisions to bring such production in conformity with
industrial licenses issued.?5.
4Paragraph 1.20
5Paragraph 1.21
132
It must be noted however, that these reports were comparing capacity created
under the licensing regime to targets specified in the five year plans. That is, ?excess
capacity? as compared to what the planners expected was needed. This is no way
meant that the existing firms had a free hand in determining their production levels
or that there was free entry into any industry. That is, even if the firm was producing
more than the amount specified on the license, it is by no means clear that it was
producing its optimal quantity given the constraints on obtaining inputs.
The constraints that the licensing regime imposed on producers were severe.
It was under the pain of imprisonment that a firm could produce above the amount
specified on its license. A firm producing plastic buckets needed permission if it
wanted to produce plastic toys instead. Availability of crucial raw materials was
at the mercy of the DGTD and was done on the basis of the capacity specified on
the license. There is also anecdotal evidence of shortages of many commodities.
For example there was a ten year waiting list for scooters and yet the owner of the
largest scooter company was hauled up about why he was producing more than his
licensed capacity 6. One of India?s largest industrialists remarked in 1969 that
?I cannot decide how much to borrow, what shares to issue, at what price, what
wages or bonus to pay and what dividend to give. I even need government permission
for the salary I pay to a senior executive.?7.
The reports that were commissioned in the 1960s and the 1970s reveal a crucial
point- whether or not the licensing Raj was able to control capacity, it was most
6Das (2000), page 175
7Das (2000), page 168
133
certainly able to control entry into an industry. In this respect licensing provided
industrialists with a great deal of protection. Evidence of this started to emerge
early in the licensing regime and the The Monopolies Inquiry Commission 1965 was
one the earliest committees established to look into this. Though the purpose of
the commission was to look at the practices of certain ?large industrial houses?, it
gave broad recommendation regarding concentration and monopoly power in Indian
industry. Based on interviews with a wide spectrum of Indian industry as well as
analysis of detailed questionnaires that were sent to firms, it stated that
?.....the requirement of law that new industries with capital over a specified
amount could not be started without a license is a formidable obstacle in the way of
new entrepreneurs freely entering the lists.? 8
Further,
?The system of controls on the shape of Industrial Licensing however necessary
from other points of views, has restricted the freedom of entry into industry and so
helped to produce concentration?9.
The Committee also mentioned the issue of high costs of production in Indian
manufacturing and averred that ? The cost of production remains high due to the
fact that top firms have not exerted themselves sufficiently......secure in the belief
that in the absence of competition from abroad there was little risk of losing their
market dominance?.10.
The Hazari Committee 1967 (Hazari (1967)) also found evidence of industri-
8Dasgupta (1965), page 7
9Dasgupta (1965), page 8
10Dasgupta (1965), page 142
134
alists pre-empting licenses. That is, a firm would send multiple applications for the
same product. This ensured that it would be granted at least some of the planned
capacity in that item, keeping out rivals. In no uncertain terms the committee
remarked that
?The obligation on all units have fixed assets more than Rs. 25 lakhs to take out
a license for new articles-applications which can be rejected out of hand on the ground
of sufficient licensed (not necessarily actual) capacity keeps at bay existing large
undertakings which might have the capacity to offer competitive products by feasible
diversification. Enterprize plus imaginative understanding of licensing formalities
thus enables the [name of large industrialist] to foreclose the market.?11
In conclusion, it seems that licensing was quite successful in providing firms
protection from entry into the industry even if the stern conditions on capacity and
output were not always implemented. Once a firm obtained a license, it held at best a
monopolistic and at worse, an oligopolistic position in the industry. Even the threat
of potential entry was low since there were reports of firms pre-empting capacity.
That is, once a firm was granted a license in a particular item it would file multiple
applications under different names so that all the ?planned production capacity?
in that item was under its control. This phenomenon has been mentioned by some
of the government reports as well as anecdotal accounts. Government officials who
scrutinized these applications report seeing as many as seven identical applications
for the same item on the same day. Even the spelling mistakes in the applications
were identical.
11Hazari (1967), paragraph 13.5
135
Thus, the licensing regime in India affected firm-level productivity and costs
through its control on both the firm?s ability and incentives to innovate, reduce
costs, adopt new technology etc. The direct controls on outputs and inputs deter-
mined ability and the indirect control of entry determined incentives. Even if the
direct controls where not implemented fully due to corruption etc, the effect of the
indirect controls on incentives was very large as evident from the fact that each and
every account of that time attributed the high costs, obsolete technology and low
productivity of Indian industry to the lack of competition, among other factors.
136
B.2 Behavior of Non-licensed Firms
In this appendix we analyze our unit-level data in greater detail. In particular
we investigate the behavior of non-licensed firms across de-licensed and licensed
industries. Our main conclusion from this analysis is that there is little evidence of
the claim that firms may choose their size (defined by assets in plant, machinery,
land and building) in anticipation of de-licensing.
B.2.1 Nature of Asset Limits
In order to show that firms were not choosing size endogenous with policy during
the 1980s, we first demonstrate some statistics. The first and important point is
whether the limit on assets in plant and machinery was binding for a non-licensed
firm. If the definition is not truly binding then the magnitude of the problem of
endogeneity is bound to be low. Table B.1 shows the utilization rates of non-licensed
firms in various percentiles below the threshold.
We see that for the entire period, the top 1% of non-licensed firms were using
on average only 80% of the rupee amount that was allowed to them. This rate drops
off sharply as the top 5% of firms use only 42.3% and the top 25% of firms use only
7.3% of the threshold.
That is, the threshold is binding only for the top 1% of firms which represent
a very small number of firms compared to the total number of non-licensed firms.
Further they represent a very small percentage of the total output of non-licensed
firms (in 1985, the top 1% of firms accounted for 6.5% of output).
137
B.2.2 Distribution of Firms Around the Threshold
Another way in which we can check whether the asset limits were binding on non-
licensed firms is by number of firms that were on or directly below the asset limit. If
the limit was binding and firms were deliberately choosing to be under the threshold,
we should see a large mass of firms on or just below the limit. In Table B.2 we
present the frequency distribution of firms in bands around the threshold. In the
first panel of the table, each column shows the number of firms that were x% below
the specified asset limit and in the second panel, the number of firms that were x%
above the asset limit.
Column 2 shows that the number of firms that were 1% below the rupee
threshold was few in number in absolute terms as well as relative to the total number
Year Mean Assets in Plant, Machinery, Utilization Rates of Permissible Limit
Land & Buildings (Rs.) within Percentiles of Threshold
99th 95th 90th
1980 2.00E+06 74.70 36.00 18.70
1981 2.09E+06 78.70 38.30 19.40
1982 2.25E+06 79.30 41.30 21.90
1983 3.58E+06 77.00 37.80 20.80
1984 3.99E+06 78.80 41.40 24.40
1985 4.05E+06 79.20 42.20 24.00
1986 4.50E+06 82.20 45.80 27.20
1987 4.76E+06 84.40 48.60 28.00
1988 5.16E+06 86.80 51.80 31.60
1989 5.31E+06 86.40 52.80 32.40
1990 9.83E+06 72.70 34.30 18.60
1991 1.06E+07 76.00 36.90 20.30
1992 1.15E+07 78.00 39.40 22.30
1993 1.23E+07 80.70 42.30 24.10
1994 1.37E+07 81.30 45.60 27.80
Table B.1: Utilization Rates for Exempt firms
138
of small firms. However one could argue that this might be because of indivisibility
of investment in plant and machinery. Thus we present the number of firms that
were within 5, 10, 20 and even 50% of the rupee threshold. Even when we take firms
that were spending more than 80% of the threshold amount (the column marked
20%) we find that they accounted for only 2% of the total number of small firms
during the 1980s. Thus there is not much evidence of bunching around the threshold
for licensing.
In the second part of the table we present the frequency distribution of firms
that were above the rupee threshold. Here we find that the frequency distribution
of firms is very similar above and below the threshold. Suppose in general firms
prefer to remain small and hence, under the licensing radar. But once a firm crosses
the threshold it has incentive to invest a lot in plant and machinery. That is, once
it is categorized as licensed by the government then it is in its interest to increase
its assets in plant and machinery as much as it wants. This means that we should
find that there aren?t too many licensed firms that are really close to the threshold.
But we see that there are as many firms in bands above the threshold as there are
below. Further, licensed firms that are 1% above the cut-off constitute on average
0.9% of the total number of licensed firms (on average over the entire period). The
corresponding figure for non-licensed firms that are 1% below the threshold is 0.04%.
If it were true that firms were choosing to remain small then the figures should be
reversed. That is, the percentage of firms directly below the threshold should be
much larger than the number of firms directly above the threshold.
Another point to note is that the rate of growth of the number of firms that
139
are 1% below the threshold (Column 2) is almost always higher than the rate of
growth of the number of firms that are 20, 50, 80% below the threshold (Column
6) as well as the rate of growth of all non-licensed firms12. This again points to
the fact that firms were not deliberately suppressing their investment in plant and
machinery in order to remain below the threshold.
B.2.3 Capital-Labor Ratio
The main issue that we want to deal with is the possibility that firms that were
Exempt from licensing (i.e. had capital below a certain defined level) might not be
targeting the absolute levels of capital. Rather they might be targeting the capital
to labor ratio which allows them to remain ?exempt? from licensing requirements
yet allow them to be competitive. Are Exempt firms choosing their capital-labor
ratio in response to the anticipation of deregulation? Analysis of data reveals that
this is not the case. In Figure B.1 below we show the distribution of capital to labor
ratio of Exempt firms in de-licensed and still licensed industries13. The dotted line
refers to the pre-reform year 1983 while the solid line refers to the post-reform year
1988. That is, we are comparing the behavior of Exempt firms across deregulated
12We also analyzed the rates of growth of assets in each band below the threshold and we find
that the rate of growth declines as the amount of assets that a firm holds declines. That is, firms
spending 80% of the threshold amount had higher rates of growth compared to firms spending 50
or 10% of the threshold amount. This means that larger non-licensed firms were growing faster.
This would not have been true if firms were deliberately trying to keep assets low.
13We do not have figures for the actual capital that a firm owns so we proxy for that using assets
in plant and machinery.
140
Year Cumulative Distribution of Firms in Bands Below the Cut-off
1% 5% 10% 20% 50% 80% 90% Total
1980 3 36 63 249 636 2298 3549 18,530
1981 5 38 74 292 727 2375 3680 18,947
1982 8 34 74 304 848 2714 4145 19,815
1983 5 33 61 281 694 2518 3833 17,960
1984 8 35 76 339 774 2692 4072 17,213
1985 3 35 73 346 793 2787 4256 17,895
1986 7 37 91 378 918 3015 4616 17,371
1987 9 46 114 413 1021 3367 5147 18,439
1988 10 66 138 457 1190 3501 5470 18,425
1989 9 84 147 509 1240 3887 5876 18,875
1990 5 20 56 252 622 2565 4060 19,389
1991 7 34 68 271 717 2739 4370 19,512
1992 8 44 86 337 884 3149 5025 21,024
1993 12 57 99 389 995 3445 5467 21,787
1994 9 54 109 429 1132 4023 6149 22,252
Year Cumulative Distribution of Firms in Bands Above the Cut-off
1% 5% 10% 20% 50% 80% 90% Total
1980 3 27 49 183 309 503 517 575
1981 3 20 46 218 350 569 590 619
1982 8 42 69 251 441 713 741 782
1983 3 31 76 225 366 522 542 581
1984 4 35 70 244 409 602 622 686
1985 10 39 88 264 452 685 711 785
1986 8 38 82 289 495 787 814 887
1987 12 52 88 346 598 931 962 1,031
1988 11 52 93 350 589 965 1005 1,192
1989 13 50 98 348 632 1021 1066 1,346
1990 7 33 54 185 316 462 473 565
1991 6 39 82 243 374 550 565 674
1992 8 45 93 258 452 679 699 867
1993 8 54 102 279 490 764 794 1,026
1994 10 65 123 351 585 914 958 1,232
Table B.2: Distribution of Firms near the Threshold for Licensing
141
and regulated industries, pre and post reforms. In case firms are anticipating dereg-
ulation, they might want to change their capital to labor ratio. However the graphs
below show that this is not true.
The distribution of capital-labor ratio is very similar across the two sets of
industries (de-licensed and licensed) in both periods (pre and post reforms). That
is, there is no systematic tendency towards using the capital to labor ratio as a
means towards preparing for anticipated deregulation.
0
.00002
.00004
.00006
.00008
0
200000 400000 600000 800000
0
200000 400000 600000 800000
Licensed De?Licensed
1983 1988
Density of Firms
Assets
Figure B.1: Distribution of Capital-Labor Ratio over Exempt Firms.
We also check for whether firms are taking the average capital to labor ratio
(implied by the nominal threshold set by the government) as the cut-off rather than
the absolute value of capital. We define ?
parenleftBigK
L
parenrightBig
t = Average
bracketleftBigK
ijt
Lijt
bracketrightBig
for all firms i that
are on or 10% around the threshold at the time t. Then we find the proportion
of exempt firms (whose absolute capital lies below the threshold) in bands around
142
Proportion of Exempt firms in Bands Around the Average KL Ratio
Year 1% 5% 10% 20% 50% 80% 90%
1980 0.49% 2.01% 2.56% 5.22% 15.29% 17.04% 6.56%
1981 0.63% 2.20% 2.59% 5.80% 17.44% 18.73% 7.91%
1982 0.60% 2.03% 2.33% 4.74% 14.58% 15.90% 6.39%
1983 0.60% 2.43% 3.01% 5.90% 18.55% 20.15% 7.57%
1984 0.44% 2.08% 2.78% 5.10% 18.18% 27.49% 11.23%
1985 0.56% 1.82% 2.47% 5.50% 15.53% 16.78% 6.28%
1986 0.37% 1.68% 2.15% 4.46% 16.91% 28.01% 13.36%
1987 0.44% 2.38% 2.74% 5.57% 17.53% 16.95% 6.37%
1988 0.57% 2.61% 2.91% 5.18% 16.49% 17.74% 6.53%
1989 0.65% 1.96% 2.94% 5.72% 17.88% 20.49% 7.67%
1990 0.37% 1.79% 2.42% 4.53% 15.95% 24.89% 13.36%
1991 0.57% 1.76% 2.88% 4.74% 16.60% 23.67% 12.52%
1992 0.52% 2.23% 2.28% 4.56% 16.73% 23.60% 12.36%
1993 0.51% 1.52% 2.49% 5.40% 16.05% 20.84% 9.20%
1994 0.40% 1.58% 1.92% 4.12% 14.52% 24.12% 13.90%
Table B.3: Distribution of Firms near the Threshold for Licensing
the average capital-labor ratio. The Table B.3 shows these figures. For example,
Column 1 shows the percentage of actually exempt firms that are either 1% below
or 1% above the average capital-labor ratio. As we can see, the proportion is very
small. Even when we allow for firms to be within 20% of the average capital to
labor ratio at the threshold, only 5% of all exempt firms (nearly 200, 000 firms) are
in that band.
143
B.3 Specification Test using Special Industries
In Table B.4 we report the results of our falsification regression using the group of
special industries in which no firm had exemption from licensing. An assumption
of our identification strategy is that the coefficient on the interaction of exemption
status and de-licensing should be insignificant for this group ofindustries. In Column
1, the coefficients are from estimation of Equation 3.3 on the group of industries
were exemption from licensing was not available (the Schedule IV and V industries).
As we see the co-efficient on the interaction of de-licensing with exemptions status
is not significant. That is, there is no exemption-based response to de-licensing in
these industries. This shows that our identification assumption holds.
However we might be worried that this is not a powerful test since the number
of observations is small as compared to the full sample. Note that the magnitude
on the interaction coefficient is large in magnitude.
Another area of concern could be the schedule IV and Schedule V industries
may not have enough variation in productivity as compared to other industries
(where exemption from licensing was allowed) and this might affect the power of
our specification test. However the coefficient of variation in log output per worker
is 17% for Schedule IV and V industries as compared with 20% for other industries.
In Column 2 of Table B.4 we present another specification test. We estimate
modified Equation 3.3 for all industries (including schedule IV and V industries).
We define Specialjt = 1 if industry j was included in Schedule IV or V in year t, 0
144
otherwise and estimate
yijts = ?0 +?j +?t +?1Dejt +?2NotExemptit +?3DejtNotExemptit (B.1)
+?4Specialjt +?5DejtNotExemptitSpecialjt +?Xijts +?ijts
That is, we include the special industries (without exemption from licensing) into
our main regression and allow them to have a different intercept and slope as com-
pared to the other industries. By pooling all industries we increase the power of
our test. The first thing to notice is that coefficient on the interaction of exemption
status and de-licensing is very small in magnitude and insignificant for the special
industries. Further we find that we accept the hypothesis that ?3 + ?5 = 0 i.e.
we accept the hypothesis that there is no variation in the performance of firms in
de-licensed industries around the threshold for licensing, for the set of industries
where this threshold was not relevant.
145
Only Special Industries All industries
NotExempt 0.637*** 0.613***
[0.094] [0.083]
De -0.001 -0.042*
[0.070] [0.025]
De*NotExempt 0.107 0.146**
[0.107] [0.065]
Special 0.072
[0.068]
Special*De*NotExempt -0.011
[0.079]
Rent-Wage Ratio -1.601*** -1.324***
[0.408] [0.288]
Constant 6.252*** 6.285***
[0.150] [0.151]
Observations 172297 206640
R-squared 0.51 0.51
Industry FE Yes Yes
(4-digit)
State FE Yes Yes
Year FE Yes Yes
F-test on restriction
?3 +?5 = 0 1.85
Table B.4: Falsification Test using Special Industries. Special refers to Schedule IV
and V industries that did not have size-based exemption from licensing.
146
B.4 Specification Test using Various Thresholds for Exemption
In this appendix we check our specification using various thresholds for exemption.
Our identification assumption implies that there should be variation in the response
of firms to de-licensing around the actual threshold for licensing. For all other hypo-
thetical thresholds, the interaction coefficient between de-licensing and exemption
status should be insignificant.
In order to conduct this specification test we use the following methodology.
The actual exemption threshold is in nominal rupees and this nominal value was
changed twice over our time period. Given the span of the data and possibility of
inflation, we can not use absolute deviations from the actual threshold to generate
our hypothetical thresholds.
We generate our hypothetical thresholds in the following way. For each year
and conditional on actual exemption status, we sort firms according to their assets
in plant, machinery, land and building. Then we take the rupee amount below which
there are 10% of the firms in that year and assume those firms are Exempt from
licensing requirements. That is, we take the 10th percentile as our first threshold.
This means that 10% of actually Exempt firms are treated as Exempt and 90% of
actually exempt firms are treated as Not Exempt. Similarly, the next threshold
is defined as the rupee amount of the 20th percentile and hence 20% of actually
Exempt firms are treated as Exempt and the other 80% are shifted into the Not
Exempt category. Thus at the 90th percentile, we treat 90% of actually Exempt
firms as exempt and 10% of actually exempt firms as Not Exempt.
147
Similarly we sort the firms above the actual threshold in each year according
to their assets and take the various percentiles as our cut-off points. Thus the first
cut-off above the actual threshold shifts 10% of the actually Not Exempt firms as
Exempt, the second cut-off shifts 20% of actually Not Exempt firms into the Exempt
category etc.
Thus we continually shift firms from the Not Exempt category into the Exempt
category as we approach the actual threshold and then go beyond it.
We define NE(x)it = 0 if firm i is in the xth percentile in year t, 1 else and es-
timate the equation below for different values of x
yijts = ?0 +?j +?t +?1Dejt +?2NE(x)it +?3DejtNE(x)it +?Xijts +?ijts (B.2)
Note that in this specification test we include the year 1980-91. This is be-
cause size-based exemption from licensing was completely removed during the 1991
reforms. That is, all firms in licensed industries were licensed and all firms in de-
licensed industries were de-licensed. This means that this robustness test is valid
only for the period in which exemption from licensing is an issue for firms.
In Table B.5 we present the coefficients, standard errors, t-statistics and z-
statistics for the interaction term DejtNE(x)it. If our identification strategy is
correct then the interaction terms should be insignificant far from the actual thresh-
old and should become more significant as we approach the threshold from either
side. As can be seen the interaction coefficient is not significant at either extremes
and rises in significance as we approach the threshold. Further as can be seen in
148
Figure B.2 (given below) where we plot the coefficient estimates and their confi-
dence interval on distance from the actual threshold, the confidence interval of the
coefficients is very large at the extremes and narrows as we approach the actual
threshold. That is, the coefficients are much more precise near the actual threshold
than away from it.
?.1
0
.1
.2
.3
0 50 100 150 200
% of firms treated as affected
Coeff Lower bound
Upper Bound
Coefficient on Interaction and Confidence Intervals
Figure B.2: Coefficient on Interaction and Confidence Intervals
149
% of actually Exempt firms Coefficient on Standard z-statistic t-statistic
treated as Not Exempt Interaction Error
90 0.13987 0.08183 1.709286 1.71
80 0.0328603 0.056206 0.584645 0.58
70 -0.0035671 0.040994 -0.08702 -0.09
60 -0.01859 0.036891 -0.50392 -0.5
50 -0.0031225 0.035701 -0.08746 -0.09
40 0.0197247 0.036726 0.537079 0.54
30 0.0452736 0.033586 1.347974 1.35
20 0.0713353 0.031008 2.300567 2.3
10 0.1040062 0.043816 2.373698 2.37
Actual threshold 0.1489872 0.08268 1.801976 1.8
110 0.1189155 0.079071 1.503915 1.5
120 0.1238649 0.077874 1.590587 1.59
130 0.1161601 0.078147 1.486438 1.49
140 0.0997598 0.077873 1.281056 1.28
150 0.113357 0.077358 1.465358 1.47
160 0.1195946 0.081337 1.470359 1.47
170 0.1067452 0.083844 1.273135 1.27
180 0.0878918 0.092328 0.951953 0.95
190 0.0876862 0.108197 0.81043 0.81
Table B.5: Coefficient on Interaction for Various Thresholds
150
B.5 Comparability of Exempt and Not Exempt Firms
Our empirical strategy is based on the differential performance of two types of
firms in response to de-licensing. That is, exempt firms (those with assets lower
than a threshold) which were not under licensing requirements provide trends in
productivity that would have occurred even without de-licensing and hence can be
used as a baseline for not exempt firms (those with assets above a threshold) which
were under licensing. However, given that the definitions of these two types of firms
are based on assets we are worried whether exempt firms even capture the basic
trends in productivity and hence whether they can be used as a baseline. That
is, are exempt (smaller) firms so different from not exempt (larger) firms that it is
unreasonable to compare the two. We run some simple checks to analyze this issue.
In Figure B.3 we show the trends in average output per worker for exempt
and not exempt firms. As we can see, exempt firms exhibit the same broad trends
in productivity over time as not exempt firms particularly in the pre-reform period
1980-85.
In Table B.6 we show the regressions results for the individual groups of ex-
empt and not exempt firms. The coefficient on the reform variable is positive and
significant for not exempt firms while it is small and negative for the exempt firms.
This is consistent with our findings in the pooled sample that not exempt firms
which were under licensing to begin with perform better after de-licensing in their
industry as compared with exempt firms in their industry. We would like to run a
simple Chow-test on these two regressions to test pool-ability. However, since we
151
6.5
7
7.5
8
8.5
9
Average Productivity
1980 1985 1990 1995
Year
Not Exempt Firms Exempt Firms
Figure B.3: Trends in Productivity-Exempt and Not Exempt Firms
Baseline Exempt Firms Not Exempt Firms
NotExempt 0.612***
[0.097]
De=1 -0.001 -0.058** 0.096*
[0.0.025] [0.022] [0.059]
De*NotExempt 0.135**
[0.07]
Constant 5.69*** 5.69*** 7.65***
[0.18] [0.18] [0.28]
Observations 178371 162285 10012
R-squared 0.522 0.49 0.584
Industry-Year (2digit) Yes Yes Yes
Industry FE (4 digit) Yes Yes Yes
State FE Yes Yes Yes
Year FE Yes Yes Yes
Table B.6: Separate Regressions for Sub-samples
152
have a large number of fixed effects in our basic model it is computationally very dif-
ficult to run the fully interacted, constrained model. But our analysis of the trends
in productivity for exempt and not exempt firms over various dimensions (location,
ownership etc) shows that both types of firms tend to more together over time. In
Figure B.4 we show the time trends in productivity for exempt and not exempt
firms for two-digit industries, in Figure B.5 for firms located in deemed backward
and not backward states; in Figure B.6 for firms with different ownership structures
and in Figure B.7 for firms with different organizational structures. These figures
reveal that the basic trends in productivity for exempt and not exempt firms are
broadly similar along several important dimensions particularly in the pre-reform
period 1980-85. Thus, exempt firms can be used as a comparison group for not
exempt firms. There are ex ante differences in the levels of productivity but these
are generally constant over time.
153
6
7
8
9
10
6
7
8
9
10
6
7
8
9
10
6
7
8
9
10
6
7
8
9
10
6
7
8
9
10
6
7
8
9
10
1980 1985 1990 1995 1980 1985 1990 1995
Cotton Textile Textile Products
Basic chemicals and chemical products Rubber, plastic, petroleum & coal products
Non?metalic mineral products Machinery, machine tools and parts
Electrical/electronic machinery, appliances etc Transport equipment & parts
Other
Wool, Silk, Synthetics, Jute, veg. Fibre Wood, Paper & Products
Basic metals & alloys & Products Food, beverage & tobacco products
Not Exempt Firms Exempt Firms
Average Productivity
Year
Figure B.4: Trends in Productivity over Industries
154
7
8
9
1980 1985 1990 19951980 1985 1990 1995
Backward States Not Backward States
Not Exempt Firms Exempt Firms
Average Productivity
Year
Figure B.5: Trends in Productivity over States
6
7
8
9
6
7
8
9
1980 1985 1990 1995
1980 1985 1990 1995
State?Owned Private Sector
Joint Sector
Not Exempt Firms Exempt Firms
Average Productivity
Year
Figure B.6: Trends in Productivity over Ownership Structure
155
4
6
8
10
4
6
8
10
4
6
8
10
4
6
8
10
1980 1985 1990 19951980 1985 1990 1995
Individual Proprietership Joint Family
Other Partnership Public Limited
Pvt Limited Pbl Corporation
Co?op Soc Others
Not Exempt Firms Exempt Firms
Average Productivity
Year
Figure B.7: Trends in Productivity over Organizational Structure
In order to get a better grasp on the problem of comparability of exempt and
not exempt firms we also run our basic regressions but for bands of firms around
the threshold. That is, we rank firms according to their assets and then regress
productivity on exemption status, de-licensing and the interaction for the top 20%
156
of exempt firms and the bottom 20% of not exempt firms14. The results of these
regressions are presented in Table B.7. Thus we see that the band of 20% of the firms
around the threshold do not show any impact of the reforms. The coefficient on the
interaction is positive and significant for 40% of the firms around the threshold and
then increases as we add more and more firms. Note that an important reason for
this is that when we select firms according to their position in the distribution of
assets, we are not able to control the industries from which these firms are selected.
This in turn means that we might not have even one firm from an industry that was
deregulated.
As a further test of comparability, we run a series of tests for the 40th percentile
of firms around the threshold (this is the first percentile where we get a significant
coefficient on the interaction of Dejt and NotExemptit). We run separate regressions
for Exempt and Not exempt firms within the 40th percentile (reported in Table B.8)
and then test whether the coefficients on key variables are the same for the two sets
of firms.
14Optimally, we would like to be able to take bands of firms around the threshold according
to how close they are to the threshold. That is, we would like to compare firms that spend 80%
of the threshold amount and 120% of the threshold amount. However, as Table B.2 shows the
distribution of firms around the threshold is quite sparse. Thus our strategy is a modified version
of the standard regression discontinuity design strategy.
157
That is, our regression equations for firms in the 40th percentile are
yijts = ?0 + ?1Dejt +?1Ratiots +?2Backwardijts +?3(Year = 1984)t
+ ?4(Year = 1991)t +?ijts for Exempt firms
yijts = ?0 + ?1Dejt +?1Ratiots +?2Backwardijts +?3(Year = 1984)t
+ ?4(Year = 1991)t +?ijts for Not Exempt firms
HereRatiots is the ratio of average cost of capital to average wages in statesin yeart,
Backwardijts = 1 if firm i was located in a backward state s in year t. The variables
(Year = 1984)t and (Year = 1991)t) are dummies for the respective years. All other
controls like the full set of year fixed effects, controls on organization, ownership etc
are included. The results for this regression are reported in Table B.8.
Now we test whether the coefficients on key variables are statistically similar
for the two types of firms. For example, for the rent-wage ratio, our null hypothesis is
that ?1 = ?1. The results for these tests are presented at the bottom of Table B.815.
We find that exempt and not exempt firms have statistically similar coefficients for
key variables like the rent-wage ratio, location of the firm (in a backward state or
not) and to the year (there were major changes in years 1984, 1985 and 1991 and we
would like to test if the average response in both types of firms is similar to these
15Our statistical program allowed us to store the separate regression results in one matrix and
test the cross-model hypotheses. The test statistic reported is the same as that for a simple Chow
test using explained sum of squares for the restricted and unrestricted models.
158
broad macroeconomic changes). Further, we also test the joint hypothesis that all
the four coefficients are similar across the models and we find that is the case. Thus,
our data shows that exempt and not exempt firms are comparable along important
dimensions16.
10th 20th 30th 40th 50th
NotExempt 0.163*** 0.238*** 0.282*** 0.338*** 0.403***
[0.035] [0.025] [0.021] [0.018] [0.016]
De=1 -0.053 -0.012 -0.029 -0.016 -0.026
[0.039] [0.028] [0.023] [0.021] [0.019]
De*NotExempt -0.047 -0.014 0.035 0.057* 0.054*
[0.061] [0.044] [0.036] [0.031] [0.028]
Constant 7.481*** 7.439*** 7.351*** 7.304*** 7.261***
[0.038] [0.028] [0.025] [0.021] [0.019]
Observations 17249 34493 51728 68975 86231
R-squared 0.45 0.4 0.36 0.35 0.34
60th 70th 80th 90th
NotExempt 0.468*** 0.545*** 0.613*** 0.683***
[0.015] [0.014] [0.014] [0.014]
De=1 -0.039** -0.031* -0.022 -0.016
[0.017] [0.016] [0.015] [0.015]
De*NotExempt 0.089*** 0.093*** 0.115*** 0.116***
[0.026] [0.024] [0.023] [0.022]
Constant 7.189*** 7.103*** 7.013*** 6.897***
[0.018] [0.017] [0.017] [0.017]
Observations 103466 120704 137933 155144
R-squared 0.35 0.35 0.37 0.41
Table B.7: Coefficient on Interaction for Various Percentiles of Firms. All regres-
sions include year, state, industry and industry-year fixed effects. All controls are
included. Robust standard errors are reported.
16We have also tried this test for ownership structure of the firm (results not reported) and we
find that we cannot reject the hypothesis that the coefficients are identical across models.
159
Pooled Not Exempt Firms Exempt Firms
NotExempt 0.348***
[0.020]
De -0.052*** 0.057 -0.057***
[0.017] [0.049] [0.017]
De*NotExempt 0.065**
[0.030]
Rent-Wage ratio -0.931*** -1.029*** -0.939***
[0.041] [0.148] [0.042]
Backward State -0.099*** -0.056* -0.100***
[0.009] [0.029] [0.010]
Constant 6.500*** 7.776*** 6.450***
[0.039] [0.204] [0.041]
Observations 67921 4813 63108
R-squared 0.49 0.5 0.48
Ho : ?1 = ?1 (Rent-Wage Ratio) F0.01(1,?) = 0.56
Ho : ?2 = ?2 (Backward State) F0.01(1,?) = 3.45
Ho : ?3 = ?3 (Year 1984) F0.01(1,?) = 2.79
Ho : ?4 = ?4 (Year 1991) F0.01(1,?) = 4.5
Ho : ?1 = ?1,?2 = ?2,?3 = ?3,?4 = ?4 F0.01(4,?) = 2.44
Table B.8: Separate Regressions for the 40th Percentile. All regressions include year
and industry fixed effects. All controls are included. Robust standard errors are
reported.
160
BIBLIOGRAPHY
Acemoglu, D., Aghion, P., and Zilibotti, F. Distance to frontier, selection and
economic growth. NBER Working Paper, (9066), 2003.
Aghion, P. and Howitt, P. A model of growth through creative destruction. Econo-
metrica, 60(2):323?351, 1992.
Aghion, P., Dewatripont, M., and Rey, P. Competition, financial discipline and
growth. Review of Economic Studies, 66:825?852, 1999.
Aghion, P., Harris, C., Howitt, P., and Vickers, J. Competition, imitation and
growth with step-by-step innovation. Review of Economic Studies, 68(3):467?492,
2001.
Aghion, P., Bloom, N., Blundel, R., Griffith, R., and Howitt, P. Competition and
innovation: An inverted u relationship. mimeo. London School of Economics, July
2003a.
Aghion, P., Burgess, R., Redding, S., and Zilibotti, F. The unequal effects of
liberalization: Theory and evidence from india. mimeo. Harvard, London School
of Economics and IIES, 2003b.
Aghion, P., Burgess, R., Redding, S., and Zilibotti, F. Entry liberalization and
inequality in industrial performance. mimeo. CEPR and London School of Eco-
nomics, Spetember 2004.
Angrist, J. and Evans, W. Schooling and labor market consequences of the 1970
state abortion reform. NBER Working Paper 5406, January 1996.
Das, D. Manufacturing productivity under varying trade regimes: India in the 1980s
and 1990s. ICRIER Working Paper 107, 2003.
Das, G. India Unbound. From Independence to the Global Information Age. Profile
Books Limited, 2nd edition, 2000.
Dasgupta, K. Report of the Monopolies Inquiry Commission. Government of India,
1965.
Delong, B. India since independence: An analytic growth narrative. In Rodrik, D.,
editor, Modern Economic Growth: Analytical Country Studies. 2001.
Djankov, S. and Murrell, P. Enterprise restructuring in transition: A quantitative
survey. Journal of Economic Literature, XL:739?792, September 2002.
Dollar, D. and Kraay, A. Trade, growth, and poverty. Finance and Development,
38(3), September 2001.
161
Duflo, E. Schooling and labor market consequences of school construction in in-
donesia: Evidence from an unusual policy experiment. NBER Working Paper
No. 7860, August 2000.
Dutt. Report of the Industrial Licensing Policy Enquiry Committee. Government of
India, 1969.
Epifani, P. Trade liberalization, firm performance and labour market outcomes in
the developing world: What can we learn from micro-level data. World Bank
Working Paper No. 3036, 2003.
Frankel, J. and Romer, D. Does trade cause growth? American Economic Review,
89(3):379?399, 1999.
Gertler, P. and Molyneaux, J. How economic development and family planning
programs combined to reduce indonesian fertility. Demography, 31(1):33?63, 1994.
Hazari, R. Industrial Planning and Licensing Policy in India. Planning Commission,
Government of India, 1967.
Hopenhayn, H. Entry, exit and firm dynamics in long-run equilibrium. Economet-
rica, 60(2):1127?1150, 1992.
Krishna, P. and Mitra, D. Trade liberalization, market discipline and productivity
growth: New evidence from india. Journal of Development Economics, 56:391?
409, 1996.
Marathe, S. S. Regulation and Development: India?s Policy Experience of Controls
over Industry. SAGE Publications, second edition, 1989.
Melitz, M. The impact of trade on intra-industry reallocations and aggregate indus-
try productivity. Econometrica, 71(6):1695?1725, November 2003.
Nickell, S. Competition and corporate performance. Journal of Political Economy,
104(4):724?746, 1996.
P. Balakrishan, K. P. and Babu, M. Trade liberalization and productivity growth
in manufacturing: Evidence from firm-level panel data. Economic and Political
Weekly, pages 3679?3682, October 7 2000.
Panagariya, A. India?s economic reforms: What has been accomplished? what
remains to be done? EDRC Policy Brief No.2, 2001.
Panagariya, A. India in the 19080?s: Weak reforms, fragile growth. Mimeo. Univer-
sity of Maryland, 2002.
Panagariya, A. India in the 1980s and 1990s: A triumph of reforms. Mimeo.
University of Maryland, 2004.
162
Ramakrishna, G. Study Group on Industrial Regulation and Procedures. Dept.
of Industrial Development, Ministry of Industry, Government of India, February
1978.
Rodrik, D. and Rodriguez, F. Trade policy and economic growth: A skeptic?s
guide to the cross-national evidence. In Bernanke, B. and Rogoff, K., editors,
Macroeconomics Annual 2000. MIT Press for NBER, 2001.
Rodrik, D. and Subramanian, A. From ?hindu growth? to productivity surge: The
mystery of the indian growth transition. IMF Staff Papers, 2004.
Rosenzweig, M. and Wolpin, K. Evaluating the effects of optimally distributed
programs: Child health and family planning interventions. American Economic
Review, 37(3):265?289, December 1988.
Sachs, J. and Warner, A. Economic reforms and the process of global integration.
Brookings Papers on Economic Activity, pages 1?118, 1995.
Schmidt, K. Managerial incentives and product market competition. The Review of
Economic Studies, 64(2):191?214, 1997.
Sivadasan, J. Barriers to Entry and Productivity: Micro-Evidence from Indian
Manufacturing Sector Reforms. PhD thesis, University of Chicago, 2003.
Tirole, J. The Theory of Industrial Organization. The MIT Press, 1988.
Topalova, P. Trade liberalization and firm productivity: The case of india. IMF
Working Paper WP/04/28, 2003.
Tybout, J. Linking trade and productivity:new research directions. The World Bank
Economic Review, 6:189?211, 1992.
Tybout, J. Manufacturing firms in developing countries: How well do they do and
why? Journal of Economic Literature, XXXVIII:11?44, March 2000.
Visaria, S. Legal reform and loan payment: The microeconomic impact of debt
recovery tribunals in india. Mimeo. Columbia University, 2004.
163